# UserOnboarding.ai > How AI products onboard new users, and how products use AI to onboard people. 38 companies, 92 screens, 101 articles, 14 questions. Collected 2026-11-06. - [Questions](https://useronboarding.ai/questions.md) ## Companies ### Onboarding for AI products - [Cursor (Anysphere)](https://useronboarding.ai/companies/cursor.md): Cursor has moved beyond a one-time install wizard to persistent setup checklists: a "Your setup" card on the web dashboard (connect GitHub/GitLab, Slack, plugins, cloud environment) and a "Getting… - [Lovable](https://useronboarding.ai/companies/lovable.md): Sign-up auto-provisions a free personal workspace, then a short dark-mode profiling step (name, then role) runs before the user lands on a home screen that is one large prompt box ("Ready to build,… - [Manus](https://useronboarding.ai/companies/manus.md): An agent product that segments users by outcome-framed role cards, shows Pro pricing with a 7-day trial before the workspace loads, and then lands on a task composer with quick-action chips and a… - [Midjourney](https://useronboarding.ai/companies/midjourney.md): A case study in onboarding through someone else's platform: through 2023 new users signed in on midjourney.com, authorized a Discord bot, joined the Discord server, passed captchas and Discord's own… - [Superhuman](https://useronboarding.ai/companies/superhuman.md): A paid-first email client, now marketed around AI, that turned its famous 30-minute 1:1 onboarding calls into a self-serve flow. A practice inbox coaches keyboard shortcuts one keypress at a time, a… - [Granola](https://useronboarding.ai/companies/granola.md): Sign in with a work Google or Microsoft account, watch a founder video that demos a live meeting, and land on a 'Get started with Granola' note built from that demo. Calendar sync means upcoming… - [Notion AI](https://useronboarding.ai/companies/notion-ai.md): No separate AI onboarding. AI opens inside the page the user is editing, with starter prompts grouped by job (write, edit, summarize). Picking one seeds a focused follow-up, and the output lands in… - [Windsurf (now Devin Desktop, Cognition)](https://useronboarding.ai/companies/windsurf.md): A desktop IDE whose first run is a short native wizard: pick a color theme, choose keybindings, optionally import settings from VS Code or Cursor, install a terminal command, then sign in. After… - [Dia (The Browser Company)](https://useronboarding.ai/companies/dia.md): An AI browser whose first-run 'Getting Started' page teaches five things: trust, the ⌘E chat shortcut, learning your patterns, history as context, and payoff by the end of the week. Setup offers… - [Otter.ai](https://useronboarding.ai/companies/otter.md): A welcome survey (work type, role, goals), a calendar step that suggests meetings for the AI notetaker to join automatically, workspace and teammate invites, and then in-product tooltips around a… - [Replit](https://useronboarding.ai/companies/replit.md): Replit's home screen is a single Agent prompt box greeting the user by name, with suggested prompts and an artifact chip (e.g. Website). Replit's own first-app guide teaches users to write an… - [v0 (Vercel)](https://useronboarding.ai/companies/v0.md): v0's landing page is the product: a "What do you want to create?" prompt box with one-click example chips and a template gallery. Per a third-party teardown, a logged-out visitor can submit a prompt… - [Claude Code (Anthropic)](https://useronboarding.ai/companies/claude-code.md): A terminal agent whose onboarding runs as a keyboard-only wizard the first time you type `claude`: an ASCII-art welcome, a text-style (theme) picker, a login-method choice that hands off to the… - [Gamma](https://useronboarding.ai/companies/gamma.md): Gamma's creation flow starts with a three-way choice — Paste in text, Generate, or Import file or URL — then a Generate screen where users pick a format, card count and language before typing a… - [GitHub Copilot (in VS Code)](https://useronboarding.ai/companies/github-copilot.md): Copilot onboards inside an editor the developer already uses: VS Code ships with a Copilot icon in the status bar, and signing in with GitHub from there silently enrolls anyone without a… - [NotebookLM (Google)](https://useronboarding.ai/companies/notebooklm.md): NotebookLM starts with the user's own sources, not a blank chat. The Add source button is the largest control in the left panel. Each upload gets an auto-generated Source Guide (summary plus key… - [Devin (Cognition)](https://useronboarding.ai/companies/devin.md): Devin's onboarding puts repository setup before the first task. Users index a repo and configure Devin's environment, then start from a 'What do you want to get done?' box with a repo selector. The… - [Figma (First Draft / Figma Make)](https://useronboarding.ai/companies/figma.md): Figma's AI features start from a prompt on the canvas instead of an empty frame. First Draft asks the user to pick one of four component libraries (from lo-fi wireframe to hi-fi site) and then… - [Meta AI (standalone app)](https://useronboarding.ai/companies/meta-ai.md): The standalone Meta AI app, launched in April 2025, is built around voice. Its home screen replaces an empty chat with an animated orb and three spoken starter options under 'Where should we… - [Magic Patterns](https://useronboarding.ai/companies/magic-patterns.md): An AI prototyping tool that teaches features during the first generation instead of before it: right after sign-up, a feature-introduction modal appears over the workspace while the first output is… ### Onboarding that uses AI - [Clay](https://useronboarding.ai/companies/clay.md): Clay skips the empty dashboard. Signup answers route new users straight into a live workbook where its Sculptor AI assistant has already suggested searches based on the user's company website, then… - [Userflow](https://useronboarding.ai/companies/userflow.md): Userflow's Smartflow builds an onboarding flow from a written purpose plus one recorded click-through of your app. The AI picks tooltips or modals and writes the copy for each step, and its Adoption… - [Webflow (AI site builder)](https://useronboarding.ai/companies/webflow-ai-site-builder.md): Webflow's AI site builder replaces the blank canvas with a prompt. You describe the site and its audience, then shape a generated multi-page draft in three steps (structure, theme, polish) before… - [Whatfix](https://useronboarding.ai/companies/whatfix.md): Whatfix, a digital adoption platform for employee software rollouts, uses AI agents on both sides. Authors prompt pop-ups, walkthroughs and display rules into existing apps, and new employees type a… - [Duolingo](https://useronboarding.ai/companies/duolingo.md): Duolingo's onboarding asks for a daily goal, then lets learners who already know some of the language take an adaptive placement test. The test uses real exercises chosen one at a time by a model to… - [Frigade](https://useronboarding.ai/companies/frigade.md): Frigade embeds an AI assistant that learns a product by using it, then onboards end users by answering in context, highlighting the next UI element step by step, or doing the setup task for them. - [HubSpot (Breeze)](https://useronboarding.ai/companies/hubspot-breeze.md): HubSpot seeds its AI context from your website. You enter a URL, Breeze crawls the site and fills a Brand Identity profile (industry, competitors, content themes, tech stack) and optionally a brand… - [Noom](https://useronboarding.ai/companies/noom.md): Noom's web onboarding is a quiz of up to 113 screens that builds a personalized weight-loss plan. It keeps redrawing a predicted goal date as the user answers and shows 'analyzing' loaders that… - [Chameleon](https://useronboarding.ai/companies/chameleon.md): Chameleon's Copilot turns a goal like 'launch our new feature' into a researched brief and then a multi-experience onboarding campaign. It uses Chameleon's benchmark data to push back on tours that… - [Pendo](https://useronboarding.ai/companies/pendo.md): Pendo's guide editor now opens with 'Start with a prompt': a product team describes the onboarding tour and who it's for, and the Guides Agent picks a layout and drafts the steps and copy for review… - [Appcues](https://useronboarding.ai/companies/appcues.md): Appcues AI lets teams build onboarding flows by chatting: describe the tour, upload a mockup or paste a page URL, and the Experience Builder agent drafts the flow in the builder for you to refine by… - [Canva](https://useronboarding.ai/companies/canva.md): Canva asks one use-case question at signup and then drops new users on a home screen built around a Canva AI prompt ('What will you design today?'). The first design can be generated from a sentence… - [Framer](https://useronboarding.ai/companies/framer.md): Framer's first-site experience can start from a prompt. Wireframer (May 2025) generates a responsive, structure-only layout, and the newer Agent panel builds and edits editable pages, copy and… - [Shopify (AI Store Builder)](https://useronboarding.ai/companies/shopify-ai-store-builder.md): New merchants describe their business in a few words and Shopify generates several complete store designs, with layout, images and copy, to pick from before touching the theme editor. - [Tandem](https://useronboarding.ai/companies/tandem.md): Tandem started as an embedded AI copilot for in-app onboarding and now sells an AI workspace for B2B implementation teams. Its agents turn customer calls into plan updates that a human approves, and… - [Userpilot](https://useronboarding.ai/companies/userpilot.md): Userpilot's agent, Lia, sits in the main nav and builds onboarding flows from a goal or from an analytics finding. Ask it to analyze activation, and it recommends a fix and offers a one-click… - [Intercom (Fin)](https://useronboarding.ai/companies/intercom-fin.md): Intercom is positioning its Fin AI agent as a 'Customer Agent' with a Success role that onboards customers after purchase: it walks them through setup step by step and suggests features they have… - [Notion (Notion Agent)](https://useronboarding.ai/companies/notion.md): Notion pitches its Agent as the way to onboard a new teammate. One prompt builds a personalized onboarding page from workspace context: a welcome, week-one priorities, a database of people to meet,… --- # Questions about onboarding for AI products Source: https://useronboarding.ai/questions ## Reported numbers - **20,000/day**: Gamma signups after rebuilding onboarding around AI, up from a few hundred ([Gamma](https://useronboarding.ai/companies/gamma.md)) - **15–20 → 3–5 min**: Replit's first-build time after its 2025 rework ([Replit](https://useronboarding.ai/companies/replit.md)) - **30% → 98%**: Superhuman panel completion once onboarding became interactive ([Superhuman](https://useronboarding.ai/companies/superhuman.md)) - **78%**: Drop-off Chameleon's benchmarks report for 5-step tours ([Chameleon](https://useronboarding.ai/companies/chameleon.md)) - **4 of 6**: AI builders that asked for signup before the first output (v0 didn't) ([v0](https://useronboarding.ai/companies/v0.md)) - **< 30 sec**: Pi, from first open to first message ## Where did onboarding go in AI products? **Into the first generation. The home screen is a prompt box, and activation happens when the first output lands.** Lovable's head of growth says she barely worked on activation at first, because it happens inside the conversation with the agent. The goal is a "wow moment": the first preview after the first prompt. The growth team now writes agent instructions to help users activate. [Lovable](https://useronboarding.ai/companies/lovable.md) Gamma spent three to four months after launch rebuilding onboarding "so that AI was part of the actual onboarding." Signups went from a few hundred a day to 20,000. [Gamma](https://useronboarding.ai/companies/gamma.md) Replit treated first-build speed as the onboarding metric and cut it from 15–20 minutes to 3–5. [Replit](https://useronboarding.ai/companies/replit.md) The home screens look alike: one big prompt box greeting you by name ([Replit](https://useronboarding.ai/companies/replit.md), [Lovable](https://useronboarding.ai/companies/lovable.md)), "What do you want to create?" ([v0](https://useronboarding.ai/companies/v0.md)), or "What can I do for you?" ([Manus](https://useronboarding.ai/companies/manus.md)). ## How do products solve the blank prompt box? **Five moves recur: starter chips, the AI speaking first, starters built from a short quiz, a feed of other people's work, and a search-style box.** The problem is named from inside the companies. OpenAI's Logan Kilpatrick says newcomers face "this blank slate". Cognition's Scott Wu says early Devin users "see the blank screen and not really know" what to do [Devin](https://useronboarding.ai/companies/devin.md). Perplexity's head of design calls prompting "the worst software experience ever". Starter chips under the box: [Canva](https://useronboarding.ai/companies/canva.md), [v0](https://useronboarding.ai/companies/v0.md), Perplexity, [Replit](https://useronboarding.ai/companies/replit.md), and Notion AI, which groups them by job (write, edit, summarize) [Notion AI](https://useronboarding.ai/companies/notion-ai.md). Manus goes further: each example opens a replay of a finished task [Manus](https://useronboarding.ai/companies/manus.md). The AI speaks first. Pi opens with "How's your day going?". Meta AI offers three spoken starters [Meta AI](https://useronboarding.ai/companies/meta-ai.md). Replika greets you by name. Starters built from a short quiz. Claude asks for your name and three topics, then offers three ideas. Picking one sends a long, well-written prompt for you, which teaches prompting by example. A feed of other people's work: Midjourney's Explore [Midjourney](https://useronboarding.ai/companies/midjourney.md), Character.ai's For You and Meta AI's Discover [Meta AI](https://useronboarding.ai/companies/meta-ai.md). Perplexity makes the box behave like search, with autocomplete, so you can type "how you would use Google". ## Should you ask questions before the first output? **Only questions whose answers visibly change what comes next. Everything else can wait or be inferred.** The spread is wide. Gemini's onboarding is 2 screens Google Gemini, Character.ai's is 3 Character.AI, and Copilot lets you chat as a guest. Replika runs about 20 steps before its paywall. Noom runs up to 113 screens, but keeps redrawing a predicted goal date so each answer feels useful [Noom](https://useronboarding.ai/companies/noom.md). The questions that pay off change the next screen. Claude's three topics become three starters. Canva's single use-case question tailors templates [Canva](https://useronboarding.ai/companies/canva.md). Clay's "How would you like to get started?" routes you into a live workbook [Clay](https://useronboarding.ai/companies/clay.md). Lovable asks name and role after signup. One teardown noted the reviewer expected to see work starting on their prompt during those screens, and it wasn't [Lovable](https://useronboarding.ai/companies/lovable.md). Microsoft Design's research on a Copilot agent argues for showing value first and personalizing later through "learned preferences, inferred context, and smart defaults, not lengthy onboarding conversations". ## Can you skip the questionnaire by reading context the user already has? **Yes, and it is the most distinctive AI-era pattern here: start from a URL, a calendar, an old app or an existing profile.** From a website URL: HubSpot Breeze crawls your site and fills industry, competitors, content themes and tech stack [HubSpot](https://useronboarding.ai/companies/hubspot-breeze.md). Clay's Sculptor suggests prospect searches from your company's site before you type anything [Clay](https://useronboarding.ai/companies/clay.md). From a calendar: Granola's upcoming meetings are "already there" after login [Granola](https://useronboarding.ai/companies/granola.md). Otter suggests meetings to auto-join [Otter.ai](https://useronboarding.ai/companies/otter.md). From the app you are leaving: Windsurf imports VS Code and Cursor settings by name [Windsurf](https://useronboarding.ai/companies/windsurf.md). Dia imports your browser data and your ChatGPT memories [Dia](https://useronboarding.ai/companies/dia.md). From an existing identity: Meta AI personalizes from linked Facebook and Instagram accounts [Meta AI](https://useronboarding.ai/companies/meta-ai.md). GitHub Copilot enrolls you in a free tier just by signing in with GitHub [GitHub Copilot](https://useronboarding.ai/companies/github-copilot.md). From one sentence about the business: Shopify generates several full stores to choose from [Shopify](https://useronboarding.ai/companies/shopify-ai-store-builder.md), and Webflow generates a multi-page draft [Webflow](https://useronboarding.ai/companies/webflow-ai-site-builder.md). The safeguard that matters: HubSpot promises "We won't overwrite anything you or your team entered manually" and asks you to review the generated brand voice [HubSpot](https://useronboarding.ai/companies/hubspot-breeze.md). ## Signup first, or value first? **The products that let people try before signing up describe it as a deliberate choice. The hard gates mostly come from fraud, age or infrastructure constraints.** Value first: Perplexity's logged-out home is a working search box, and signup is pitched as "unlock the full potential". v0 starts generating before you sign up; a teardown found 4 of 6 AI builders asked for signup up front, and v0 didn't [v0](https://useronboarding.ai/companies/v0.md). Copilot lets guests chat and ties sign-in to "longer conversations". Gate first: ChatGPT's 2023 iOS flow required name, birthday, phone verification and a puzzle before the first message. Midjourney's Discord era took 13 screens and two captchas [Midjourney](https://useronboarding.ai/companies/midjourney.md). Claude Code assumes you already have a paid account [Claude Code](https://useronboarding.ai/companies/claude-code.md). Granola blocks free email providers [Granola](https://useronboarding.ai/companies/granola.md). A middle path: borrow an account people already have. GitHub sign-in enrolls you in Copilot Free [GitHub Copilot](https://useronboarding.ai/companies/github-copilot.md). Meta AI uses Facebook or Instagram [Meta AI](https://useronboarding.ai/companies/meta-ai.md). ## Where does pricing show up? **Earlier than in classic SaaS, often before the first message, because every output costs money. The better flows teach the meter along with the price.** Mid-onboarding plan pickers: Claude shows Free, Pro and Max before the first chat, labeled "Stay on Free plan". Perplexity shows plans right after signup. Manus [Manus](https://useronboarding.ai/companies/manus.md) and Grammarly offer 7-day trials inside setup. Replika puts a paywall after a long quiz. After value: Gemini's upgrade comes later in use. Character.ai has a soft paywall after onboarding. Pi has none. Credits as onboarding. Lovable starts at 5 daily credits [Lovable](https://useronboarding.ai/companies/lovable.md). Manus shows a balance that refills at midnight [Manus](https://useronboarding.ai/companies/manus.md). Magic Patterns warns "Almost at your limit" during the first session [Magic Patterns](https://useronboarding.ai/companies/magic-patterns.md). Clay pays out 100 credits for your first enrichment, which works as a reward because every action costs credits [Clay](https://useronboarding.ai/companies/clay.md). The Manus teardown puts it plainly: "If usage is metered, teach the meter during onboarding." [Manus](https://useronboarding.ai/companies/manus.md) ## Should the first output be finished or a draft? **A draft you can steer. The creative and site builders all add a checkpoint before the expensive generation.** Gamma's main button says "Generate outline", not "Generate deck" [Gamma](https://useronboarding.ai/companies/gamma.md). Figma renamed Make Designs to First Draft so the output would read as a starting point [Figma](https://useronboarding.ai/companies/figma.md). Framer's Wireframer generates "structure, not style" [Framer](https://useronboarding.ai/companies/framer.md). Webflow walks you through structure, then theme, then polish [Webflow](https://useronboarding.ai/companies/webflow-ai-site-builder.md). Choice instead of one answer: Shopify generates several stores to pick from [Shopify](https://useronboarding.ai/companies/shopify-ai-store-builder.md). Replit added a flow to "visualize what Agent would build before committing" [Replit](https://useronboarding.ai/companies/replit.md). Lovable has a Plan mode that writes a plan without changing code [Lovable](https://useronboarding.ai/companies/lovable.md). The output is real, editable material: Figma uses real library components [Figma](https://useronboarding.ai/companies/figma.md), Framer produces real layers [Framer](https://useronboarding.ai/companies/framer.md) and Webflow a design system [Webflow](https://useronboarding.ai/companies/webflow-ai-site-builder.md). A picture of a design would dead-end. ## Do product tours still work? **Tours that describe features are fading. Tours that make you do the thing still work, and tips have moved into the time spent waiting on generation.** Chameleon's benchmarks: 5-step tours see 78% drop-off, and 91% of users dismiss the first onboarding step [Chameleon](https://useronboarding.ai/companies/chameleon.md). But Superhuman's interactive panels went from 30% to 98% completion. Its practice inbox won't advance until you press the real key [Superhuman](https://useronboarding.ai/companies/superhuman.md). Its growth lead's rule is do > show > tell. Teaching moves to the wait: Magic Patterns shows a feature modal and rotating tips while the first output renders [Magic Patterns](https://useronboarding.ai/companies/magic-patterns.md). Teaching moves to the moment of use: ChatGPT's voice intro appears the first time you open voice. NN/g praises Perplexity's hover tooltips over upfront tutorials. Notion AI lives in the page you are already editing [Notion AI](https://useronboarding.ai/companies/notion-ai.md). Checklists persist, and they break when completion detection misses. Cursor's forums show users stuck at 3/4 after finishing a step [Cursor](https://useronboarding.ai/companies/cursor.md). ## How do products teach people to prompt? **By example more than by instruction: seeded prompts, replays and small first tasks.** Claude's starter ideas send a long prompt that tells the model to ask follow-up questions and use tools, so the first message is a model prompt. Replit's first-app guide teaches one outcome plus success criteria, and postpones login, data and payments [Replit](https://useronboarding.ai/companies/replit.md). Vercel teaches a three-part prompt (product surface, context of use, constraints and taste), which in its test came out 19 seconds faster with 152 fewer lines [v0](https://useronboarding.ai/companies/v0.md). Start small. Devin's onboarding pushes repo setup and "a couple one pointer tasks" first: "treat Devin like your new junior engineer" [Devin](https://useronboarding.ai/companies/devin.md). Claude Code's docs suggest questions like "what does this project do?" before any edit [Claude Code](https://useronboarding.ai/companies/claude-code.md). Package the prompt so there is nothing to learn: Gemini's Gems Google Gemini, ChatGPT's GPTs ChatGPT, and Dia's preloaded Skills [Dia](https://useronboarding.ai/companies/dia.md). ## How do AI products set expectations about mistakes and data? **Up front and briefly: one screen of cautions, an "AI can make mistakes" line near the input, and approval before the AI changes anything.** ChatGPT's welcome screen went from three cautions to two: accuracy, and sensitive info used for training. Claude opens with "Data, safety, and you". Dia puts "Built for your trust" first of its five things to know [Dia](https://useronboarding.ai/companies/dia.md). Microsoft Design treats the Copilot zero state as education, because the first encounter with an LLM is "potentially intimidating". It shows capabilities, limits and a "Check for mistakes" notice from the first visit. When AI acts on your data, it asks first. Tandem shows "Nothing is applied yet" with a quote from the call explaining each change [Tandem](https://useronboarding.ai/companies/tandem.md). Frigade fills in the invite but leaves the final click to the user [Frigade](https://useronboarding.ai/companies/frigade.md). Userflow's action flows run only after a click [Userflow](https://useronboarding.ai/companies/userflow.md). Age gates are standard for companion apps: Character.ai asks for a birthday Character.AI, and Replika asks for an age range. ## What counts as activated? **Teams that name it define a single concrete first success, and several say they got results by pushing one thing hard.** Named activation moments: Lovable's first preview [Lovable](https://useronboarding.ai/companies/lovable.md), Clay's first enrichment run on a live table [Clay](https://useronboarding.ai/companies/clay.md), Devin's repo set up plus a first small task [Devin](https://useronboarding.ai/companies/devin.md), and Gamma's "magic in the first 30 seconds" [Gamma](https://useronboarding.ai/companies/gamma.md). One thing, not five. Gamma's "one egg" rule: show a new user one idea, not four or five features, and treat them as "selfish, vain, and lazy" [Gamma](https://useronboarding.ai/companies/gamma.md). Superhuman's self-serve activation rose from 40% to 50% after onboarding pushed one opinionated feature [Superhuman](https://useronboarding.ai/companies/superhuman.md). Humans set the bar. Superhuman's 1:1 onboarding calls led 65% of customers to fully switch, more than double the self-serve rate, and the practice inbox was built to close that gap [Superhuman](https://useronboarding.ai/companies/superhuman.md). Figma has a team called "Blockers" whose fixes showed up as activation gains [Figma](https://useronboarding.ai/companies/figma.md). ## What are onboarding-tool vendors doing with AI? **They all ship "describe the goal and we'll draft the flow." The frontier is an agent that does the setup for the user, which would make the tour unnecessary.** Prompt-to-flow builders: Appcues builds from a request, a mockup or a URL [Appcues](https://useronboarding.ai/companies/appcues.md). Pendo's editor opens with "Start with a prompt" [Pendo](https://useronboarding.ai/companies/pendo.md). Userflow records one click-through of your app [Userflow](https://useronboarding.ai/companies/userflow.md). Whatfix turns plain-language targeting rules into config [Whatfix](https://useronboarding.ai/companies/whatfix.md). Analytics to fix: Userpilot's Lia answers "Analyze my activation" with a funnel, a recommendation and a Create Onboarding Flow button [Userpilot](https://useronboarding.ai/companies/userpilot.md). Chameleon's Copilot pushes back on tours its benchmarks say users dismiss [Chameleon](https://useronboarding.ai/companies/chameleon.md). Agents that act: Frigade's assistant learns your product by using it and can do the setup task [Frigade](https://useronboarding.ai/companies/frigade.md). Userflow's Adoption Agent answers with a matching walkthrough [Userflow](https://useronboarding.ai/companies/userflow.md). Intercom is adding an onboarding role to Fin [Intercom](https://useronboarding.ai/companies/intercom-fin.md). One vendor already left: Tandem pivoted from an in-app copilot to a workspace for implementation teams [Tandem](https://useronboarding.ai/companies/tandem.md). ## How do voice-first products onboard? **They open on voice, with one voice shown at a time and a short description of its tone.** ChatGPT's advanced voice has its own intro sheet, then a carousel ("Ember, confident and optimistic"). Gemini Live shows one voice at a time with a tone line. Pi offers eight voices inside the conversation. Meta AI's home is an animated orb and three spoken starters, with typing as the fallback [Meta AI](https://useronboarding.ai/companies/meta-ai.md). ## What can't this data tell us? **Mostly what worked. Results numbers are rare, and the screens show what a product does, not how well it converts.** Only a handful of teams report outcomes: Gamma, Replit, Superhuman, Chameleon's benchmarks and Vercel's prompt test. Everything else shows what a product does, not whether it works. Some captures are dated: Duolingo's are from 2019 [Duolingo](https://useronboarding.ai/companies/duolingo.md). Dia's images are illustrations from its help page, not real screens [Dia](https://useronboarding.ai/companies/dia.md). Noom is included as a reference pattern; none of its sources call its personalization AI [Noom](https://useronboarding.ai/companies/noom.md). Missing entirely: Grok and Poe had no public onboarding screenshots. Creative tools like Runway, ElevenLabs and Suno haven't been researched yet. --- # Cursor (Anysphere) Source: https://useronboarding.ai/companies/cursor - Category: Onboarding for AI products - Type: Dev tools - Product: https://cursor.com Cursor has moved beyond a one-time install wizard to persistent setup checklists: a "Your setup" card on the web dashboard (connect GitHub/GitLab, Slack, plugins, cloud environment) and a "Getting Started" progress card in the desktop sidebar, plus a cursor.com/onboard page for picking a repo so cloud agents can run and test it. Public forum threads show the checklists are sticky when completion detection misses. ## Notes - The dashboard "Your setup" checklist (April 2026 forum screenshot) has four items — Connect GitHub or GitLab, Connect Slack, Extend Cursor with plugins, and "Set up your cloud environment for faster, parallelizable agents everywhere" with a Set up button — shown directly above an AI Line Edits contribution heatmap and streak stats. - A Cursor staffer confirmed the checklist "doesn't detect cloud agent usage for individual accounts (only team-based setups)", leaving users stuck at 3/4 even though the feature worked; a second user reported the "Getting started" flow wouldn't disappear after finishing environment setup. - The desktop app sidebar carries a "Getting Started 2/3" card with a progress ring and a single next action ("Connect Slack"); an August 2026 thread shows it staying at 2/3 after Slack was linked until the app was restarted. - cursor.com/onboard frames repo setup around the payoff — headline "Get demos, not diffs" — asking the user to select a repository and provide environment secrets so agents can test their work; one thread shows the dropdown empty ("No repositories available") for a ~300-project GitLab org. - An older 2024 thread shows the original IDE welcome flow tracked progress too: users saw "20% done" in the WELCOME section and could not re-trigger onboarding after skipping steps, even after reinstalling and clearing ~/Library files. - Cursor designer John Bai announced in a public LinkedIn post (2026) that Cursor shipped in-product onboarding tips. The post gives no design rationale. A commenter pointed to the gap it fills: users find useful features months after they start (LinkedIn). ## Patterns setup checklist, progress indicator, integration connect steps, benefit-led headline, persistent getting-started card ## Screens - ![Dashboard "Your setup" checklist at 3/4 (GitHub, Slack, plugins done; cloud environment pending) above the AI Line Edits heatmap](https://useronboarding.ai/images/cursor-1.jpg) Dashboard "Your setup" checklist at 3/4 (GitHub, Slack, plugins done; cloud environment pending) above the AI Line Edits heatmap (source: https://forum.cursor.com/t/stuck-on-cloud-environment-setup-3-4-new-your-setup-checklist-won-t-finish/158678) - ![Desktop app sidebar with a "Getting Started 2/3" card and Connect Slack button](https://useronboarding.ai/images/cursor-3.jpg) Desktop app sidebar with a "Getting Started 2/3" card and Connect Slack button (source: https://forum.cursor.com/t/slack-and-desktop-getting-started-still-show-connect-link-after-slack-is-already-linked/168215) - ![cursor.com/onboard: "Get demos, not diffs" repo picker showing "No repositories available" with Refresh / Add Repositories](https://useronboarding.ai/images/cursor-2.jpg) cursor.com/onboard: "Get demos, not diffs" repo picker showing "No repositories available" with Refresh / Add Repositories (source: https://forum.cursor.com/t/cursor-onboard-not-showing-repositories/160417) ## Articles - [Shipped onboarding tips in Cursor (John Bai)](https://www.linkedin.com/posts/john-bai_shipped-onboarding-tips-in-cursor-if-you-activity-7463354101674835968-R0CW): LinkedIn post by John Bai, Cursor designer (by the product team) - [Stuck on cloud environment setup (3/4) — new "Your setup" checklist won't finish](https://forum.cursor.com/t/stuck-on-cloud-environment-setup-3-4-new-your-setup-checklist-won-t-finish/158678): Cursor Community Forum - [Slack and desktop Getting Started still show Connect/Link after Slack is already linked](https://forum.cursor.com/t/slack-and-desktop-getting-started-still-show-connect-link-after-slack-is-already-linked/168215): Cursor Community Forum - [Cursor onboard not showing repositories](https://forum.cursor.com/t/cursor-onboard-not-showing-repositories/160417): Cursor Community Forum - [Re-trigger the onboarding experience](https://forum.cursor.com/t/re-trigger-the-onboarding-experience/21086): Cursor Community Forum ## Cited in - [Do product tours still work?](https://useronboarding.ai/questions#tours) --- # Clay Source: https://useronboarding.ai/companies/clay - Category: Onboarding that uses AI - Type: Productivity - Product: https://www.clay.com/sculptor Clay skips the empty dashboard. Signup answers route new users straight into a live workbook where its Sculptor AI assistant has already suggested searches based on the user's company website, then it suggests an enrichment and pays out credits when the user runs it. ## Notes - Signup collects workspace name, team type, CRM, tools and attribution via click-to-select tiles; the last question, 'How would you like to get started?', offers Find and enrich people/accounts, Generate Pre-Meeting Notes, AI Outbound Messaging and Automate Account Research (Supademo teardown). - The workbook opens with 'I'm Sculptor, your AI assistant for Clay' and three 'Suggested searches for supademo.com based on your business context' (e.g. 'Customer Success Managers at SaaS firms'), with a link showing what context was used. - After the first search, Clay highlights a suggested enrichment column ('Work Email', ~2.9 credits/row, 'Save and run 10 rows'). - Approving it triggers 'You've earned 100 credits for creating your first waterfall!'; the teardown notes the reward is real because every action in Clay costs credits. - The teardown defines activation as completing the first enrichment run on a live prospect table. ## Patterns url-to-workspace, empty-state prompts, use-case routing, ai-suggested next action, activation rewards ## Screens - !['How would you like to get started?' use-case picker that routes into a first project](https://useronboarding.ai/images/clay-1.jpg) 'How would you like to get started?' use-case picker that routes into a first project (source: https://supademo.com/user-flow-examples/clay) - ![Sculptor's suggested searches generated from the signup company's website/business context](https://useronboarding.ai/images/clay-2.jpg) Sculptor's suggested searches generated from the signup company's website/business context (source: https://supademo.com/user-flow-examples/clay) - ![Suggested 'Work Email' enrichment with per-row credit cost and 'Save and run 10 rows'](https://useronboarding.ai/images/clay-3.jpg) Suggested 'Work Email' enrichment with per-row credit cost and 'Save and run 10 rows' (source: https://supademo.com/user-flow-examples/clay) - !['You've earned 100 credits for creating your first waterfall!' activation reward](https://useronboarding.ai/images/clay-4.jpg) 'You've earned 100 credits for creating your first waterfall!' activation reward (source: https://supademo.com/user-flow-examples/clay) ## Articles - [Sculptor](https://www.clay.com/sculptor): Clay (by the product team) - [Sculptor - Clay Docs](https://university.clay.com/docs/sculptor): Clay University (by the product team) - [Clay Onboarding Flow: A Screen-by-Screen Teardown](https://supademo.com/user-flow-examples/clay): Supademo ## Cited in - [Should you ask questions before the first output?](https://useronboarding.ai/questions#ask-or-infer) - [Can you skip the questionnaire by reading context the user already has?](https://useronboarding.ai/questions#context-not-questions) - [Where does pricing show up?](https://useronboarding.ai/questions#pricing) - [What counts as activated?](https://useronboarding.ai/questions#activation) --- # Lovable Source: https://useronboarding.ai/companies/lovable - Category: Onboarding for AI products - Type: Dev tools - Product: https://lovable.dev Sign-up auto-provisions a free personal workspace, then a short dark-mode profiling step (name, then role) runs before the user lands on a home screen that is one large prompt box ("Ready to build, ?") with a Build/Chat/Plan mode picker. ## Notes - Per the account docs, Lovable "automatically creates your first personal workspace on the free plan" right after sign-up (email/password, Google, GitHub, Apple, or SAML/OIDC SSO). - Growthmates' teardown (Kate Syuma) found Lovable still collects profiling data after sign-up — only name and role — via one-question screens with a 4-dot progress indicator; role options are Founder, Product, Designer, Engineer, Consultant, Marketing/Sales, Operations, Other. She noted she expected to see work in progress on her prompt at that point, but it was missing. - Free workspaces start with "5 daily build credits (up to 30 per month)"; the docs point new users to a Quick start estimated at 10 minutes. - Home screen headline greets by first name ("Ready to build, Ines?") over a full-bleed gradient; the placeholder rotates example intents ("Ask Lovable to create a blog about…", "…a prototype…"). - The mode dropdown next to send offers Build, Chat ("think through an idea" outside a project) and Plan ("investigates and writes a plan without changing code") — letting hesitant first-timers start without generating code. - The + menu inside the prompt exposes Attach, Design (templates and design systems) and Connectors, so templates live inside the composer rather than a separate gallery step; a Templates tab also sits below the fold next to My projects. - Growth hooks are on the first screen: a "New: Lovable is now on Telegram" pill above the headline, and sidebar cards for "Share Lovable — 100 credits per paid referral" and "Upgrade to Pro". - Head of Growth Elena Verna says she barely worked on activation at first. At Lovable, activation happens inside the conversation with the agent, and the core agent team "thinks so much about that first generation". A better agent improves the first build and the hundredth at once (Lenny's Podcast, 2025-12-18, ~21:50-23:32). - Verna says the goal is a "wow moment", not an aha moment: "that first preview generation after your first prompt". It won't be complete, but users think "This is possible?" and keep building. The growth team now writes agent instructions "in order for customers to activate better" (Lenny's Podcast, ~22:00 and ~53:00). ## Patterns role question, single prompt box home, personalized greeting, rotating placeholder examples, plan mode before build, templates inside composer, auto-created workspace, referral credits, first generation as activation ## Screens - ![Post-signup profiling: "What's your name?" and "Which role fits you best?" (8 role tiles) with step dots](https://useronboarding.ai/images/lovable-3.jpg) Post-signup profiling: "What's your name?" and "Which role fits you best?" (8 role tiles) with step dots (source: https://www.growthmates.news/p/your-onboarding-should-become-invisible) - ![Home screen on a free workspace: "Ready to build, Ines?" prompt box with Build mode, Telegram announcement pill, referral and upgrade cards](https://useronboarding.ai/images/lovable-1.jpg) Home screen on a free workspace: "Ready to build, Ines?" prompt box with Build mode, Telegram announcement pill, referral and upgrade cards (source: https://docs.lovable.dev/introduction/welcome) - ![Prompt box + menu open: Attach, Design and Connectors options; Templates tab below](https://useronboarding.ai/images/lovable-2.jpg) Prompt box + menu open: Attach, Design and Connectors options; Templates tab below (source: https://docs.lovable.dev/introduction/dashboard-overview) ## Articles - [Create an account](https://docs.lovable.dev/introduction/create-an-account): Lovable Docs (by the product team) - [Dashboard overview](https://docs.lovable.dev/introduction/dashboard-overview): Lovable Docs (by the product team) - [The new AI growth playbook for 2026 | How Lovable hit $200M ARR in one year (Elena Verna)](https://www.youtube.com/watch?v=6qAB6aUMIeA): Lenny's Podcast (YouTube, 2025-12-18) (by the product team) - [Your onboarding should become invisible](https://www.growthmates.news/p/your-onboarding-should-become-invisible): Growthmates (Kate Syuma) ## Cited in - [Where did onboarding go in AI products?](https://useronboarding.ai/questions#where-did-onboarding-go) - [Should you ask questions before the first output?](https://useronboarding.ai/questions#ask-or-infer) - [Where does pricing show up?](https://useronboarding.ai/questions#pricing) - [Should the first output be finished or a draft?](https://useronboarding.ai/questions#draft-first) - [What counts as activated?](https://useronboarding.ai/questions#activation) --- # Manus Source: https://useronboarding.ai/companies/manus - Category: Onboarding for AI products - Type: Assistants - Product: https://manus.im An agent product that segments users by outcome-framed role cards, shows Pro pricing with a 7-day trial before the workspace loads, and then lands on a task composer with quick-action chips and a visible credit balance. ## Notes - 'Which best describes your work?' shows nine role cards (Business Owner, Growth, E-commerce, etc.) plus Others, each with a one-line outcome, and a Skip link top-right (AI UX Playground). - 'Upgrade to Manus Pro' appears mid-onboarding with three tiers, the middle '7-Day Free' tier highlighted, and credit-volume dropdowns on higher tiers. - Post-onboarding home is 'What can I do for you?' with chips (Create slides, Build website, Design) plus a desktop-app download banner and a 'Learn to build, earn free credits' banner. - A credit pill opens a popover with the free balance and a daily refresh to 300 credits at 00:00, plus View usage and Upgrade. - Feature-announcement modals (e.g. 'Introducing Shopify On Manus') teach new capabilities without blocking the first task. - Teardown takeaway: 'If usage is metered, teach the meter during onboarding.' - NN/g uses Manus as its example of use-case prompt suggestions: category tabs of example tasks under the input, each opening a full replay of a finished task, so new users see what the agent can do before they type (NN/g, 2025-04-25). ## Patterns role picker, outcome-framed options, paywall during onboarding, composer as home, credit meter, feature announcement modal ## Screens - !['Which best describes your work?' nine outcome-framed role cards with Skip](https://useronboarding.ai/images/manus-1.jpg) 'Which best describes your work?' nine outcome-framed role cards with Skip (source: https://aiuxplayground.com/teardowns/manus/onboarding) - !['Upgrade to Manus Pro' with the 7-day free tier highlighted](https://useronboarding.ai/images/manus-2.jpg) 'Upgrade to Manus Pro' with the 7-day free tier highlighted (source: https://aiuxplayground.com/teardowns/manus/onboarding) - ![First home: 'What can I do for you?' composer, quick-action chips, desktop app promo](https://useronboarding.ai/images/manus-3.jpg) First home: 'What can I do for you?' composer, quick-action chips, desktop app promo (source: https://aiuxplayground.com/teardowns/manus/onboarding) - ![Credits popover: free balance and daily refresh at midnight](https://useronboarding.ai/images/manus-4.jpg) Credits popover: free balance and daily refresh at midnight (source: https://aiuxplayground.com/teardowns/manus/onboarding) - !['Introducing Shopify On Manus' feature modal on first run](https://useronboarding.ai/images/manus-5.jpg) 'Introducing Shopify On Manus' feature modal on first run (source: https://aiuxplayground.com/teardowns/manus/onboarding) ## Articles - [Manus onboarding UX: role segmentation & Pro upsell](https://aiuxplayground.com/teardowns/manus/onboarding): AI UX Playground - [Prompt Suggestions](https://www.nngroup.com/articles/prompt-suggestions/): Nielsen Norman Group (Kate Moran, Tim Neusesser) ## Cited in - [Where did onboarding go in AI products?](https://useronboarding.ai/questions#where-did-onboarding-go) - [How do products solve the blank prompt box?](https://useronboarding.ai/questions#blank-box) - [Where does pricing show up?](https://useronboarding.ai/questions#pricing) --- # Midjourney Source: https://useronboarding.ai/companies/midjourney - Category: Onboarding for AI products - Type: Creative - Product: https://www.midjourney.com A case study in onboarding through someone else's platform: through 2023 new users signed in on midjourney.com, authorized a Discord bot, joined the Discord server, passed captchas and Discord's own community onboarding, then added the bot to a server before generating anything. The 2024 web app replaced this with an Explore feed of community images and an "Imagine…" prompt bar, initially gated behind prior Discord usage. ## Notes - Faqprime's 2023 walkthrough counts 13 screens from landing page to "All set!": Sign In → Discord OAuth "Authorize" (Midjourney Bot requests username, email, server joins, etc.) → "Join the Discord to start creating!" → Accept Invite → captcha → Discord's "Start Onboarding" gate → find Midjourney Bot → Add to Server → select server → Continue → authorize → second captcha. - The pre-Discord web dashboard's empty state was just a "Join the Discord to start creating!" link and "When you generate an image, it will appear here" — the website could not create, only display. - Two separate captcha checks and a Discord-native "You must complete Onboarding to view this community" gate sit between sign-in and the first prompt. - How-To Geek (May 19, 2024) reports the new web interface was first limited to users who had generated 1,000 images on Discord, then "lowered way down to only 100 images" — access to the easier onboarding was earned through the harder one. - The web alpha opens on Explore (Random / Hot / Top Day / Likes tabs, a Randomize button, and a "Search prompts" box) with the "Imagine…" bar pinned at the top, so the first screen is a gallery of what's possible plus the input to try it; sidebar also offers Rate Images and Rate Ideas. - Per How-To Geek, generation settings (size, aesthetics, model) hide behind the sliders icon next to the prompt rather than being asked up front. ## Patterns third-party platform onboarding, OAuth authorize, captcha, community gallery as home, usage-gated access, empty state with next step ## Screens - ![Discord OAuth screen asking the user to authorize Midjourney Bot's permissions](https://useronboarding.ai/images/midjourney-1.jpg) Discord OAuth screen asking the user to authorize Midjourney Bot's permissions (source: https://faqprime.com/en/onboarding-on-midjourney/) - ![Web dashboard empty state: "Join the Discord to start creating!"](https://useronboarding.ai/images/midjourney-2.jpg) Web dashboard empty state: "Join the Discord to start creating!" (source: https://faqprime.com/en/onboarding-on-midjourney/) - ![Inside the Midjourney Discord server: "You must complete Onboarding to view this community" with a Start Onboarding button](https://useronboarding.ai/images/midjourney-3.jpg) Inside the Midjourney Discord server: "You must complete Onboarding to view this community" with a Start Onboarding button (source: https://faqprime.com/en/onboarding-on-midjourney/) - ![Adding Midjourney Bot to the user's own Discord server (server select step)](https://useronboarding.ai/images/midjourney-4.jpg) Adding Midjourney Bot to the user's own Discord server (server select step) (source: https://faqprime.com/en/onboarding-on-midjourney/) - ![Midjourney web alpha: Explore feed with "Imagine…" bar, Search prompts, and Random/Hot/Top Day/Likes tabs](https://useronboarding.ai/images/midjourney-5.jpg) Midjourney web alpha: Explore feed with "Imagine…" bar, Search prompts, and Random/Hot/Top Day/Likes tabs (source: https://www.howtogeek.com/midjourney-now-has-web-interface-heres-how-to-use-it/) ## Articles - [Onboarding on Midjourney](https://faqprime.com/en/onboarding-on-midjourney/): Faqprime (user flows) - [Midjourney Now Has Web Interface, Here's How to Use It](https://www.howtogeek.com/midjourney-now-has-web-interface-heres-how-to-use-it/): How-To Geek ## Cited in - [How do products solve the blank prompt box?](https://useronboarding.ai/questions#blank-box) - [Signup first, or value first?](https://useronboarding.ai/questions#signup-timing) --- # Superhuman Source: https://useronboarding.ai/companies/superhuman - Category: Onboarding for AI products - Type: Productivity - Product: https://superhuman.com A paid-first email client, now marketed around AI, that turned its famous 30-minute 1:1 onboarding calls into a self-serve flow. A practice inbox coaches keyboard shortcuts one keypress at a time, a Get Started checklist (which now includes 'Activate Superhuman AI') follows, and a 'Get me to Zero' step clears old mail. ## Notes - First Round (interview with Gaurav Vohra, Apr 2025): 1:1 sessions ran 30 minutes (once up to 90); over 65% of human-onboarded customers fully switched their email, more than double the self-serve rate; the synthetic practice inbox carried those lessons into the product. - In-product welcome: 'Superhuman is designed to get you through your inbox twice as fast. Let's master some shortcuts…' with an 'ONWARDS — hit enter' button and a progress bar (First Round). - The Get Started checklist ('5 steps left') lists Clean up old email, Turn on read statuses, Activate Superhuman AI, Customize your calendar, Get Superhuman for iOS, and the keyboard hint bar stays pinned at the bottom (First Round). - 'Welcome to your Inbox. Now let's get you to Inbox Zero!' moves mail older than a chosen age ('1 week (most popular)') to Done, with 'Get me to Zero!' or 'Do this later' (First Round). - Page Flows (desktop web, about 2022) shows plan choice and checkout before product use, then 'Learn Superhuman in 5 minutes — We'll use a practice inbox to master the basics'. - The practice inbox gates each step on a real keypress ('Hit J to go down. Hit K to go up', then 'Hit enter to open') and ends on a full-screen 'Congratulations! You hit Inbox Zero.' - Growth lead Gaurav Vohra says in-product onboarding should be opinionated ("there is likely a best way"), interruptive ("If an experience is tucked away, it will be ignored") and interactive (do > show > tell) (Gaurav Vohra, 2025-04-23). - Reported results: self-serve activation rose from 40% to 50% after onboarding pushed a single opinionated feature. The full-screen 'Get Me To Zero' interrupt reached 57% opt-in. Completion of the interactive onboarding panels jumped from 30% to 98% (Gaurav Vohra). - The self-serve checklist lists 'Activate Superhuman AI' next to 'Clean up old email' and 'Get Superhuman for iOS', so the AI features are one step on the path, not the first screen. The essay says little else about AI (Gaurav Vohra). ## Patterns human concierge onboarding, practice sandbox, keypress-gated tutorial, checklist, bulk cleanup to zero, celebration moment, paywall before product, interruptive full-screen panels, learn by doing, onboarding checklist, human-led to self-serve ## Screens - !['Welcome to Superhuman … Let's master some shortcuts' with an ONWARDS / hit enter button](https://useronboarding.ai/images/superhuman-1.jpg) 'Welcome to Superhuman … Let's master some shortcuts' with an ONWARDS / hit enter button (source: https://review.firstround.com/superhuman-onboarding-playbook/) - ![Get Started checklist including 'Activate Superhuman AI', with the shortcut hint bar](https://useronboarding.ai/images/superhuman-2.jpg) Get Started checklist including 'Activate Superhuman AI', with the shortcut hint bar (source: https://review.firstround.com/superhuman-onboarding-playbook/) - !['Welcome to your Inbox — Get me to Zero!' cleanup of email older than 1 week](https://useronboarding.ai/images/superhuman-3.jpg) 'Welcome to your Inbox — Get me to Zero!' cleanup of email older than 1 week (source: https://review.firstround.com/superhuman-onboarding-playbook/) ## Articles - [How to Build and Scale Onboarding (Superhuman's Onboarding Playbook)](https://review.firstround.com/superhuman-onboarding-playbook/): First Round Review (interview with Gaurav Vohra, Superhuman) (by the product team) - [Obsessing Over Onboarding for 10+ Years (Superhuman's onboarding playbook)](https://substack.gauravvohra.com/p/obsessing-over-onboarding-for-10): Gaurav Vohra's Substack (also published by First Round Review) (by the product team) - [Superhuman's Secret 1-on-1 Onboarding Revealed](https://growth.design/case-studies/superhuman-user-onboarding): growth.design - [Onboarding on Superhuman (desktop web)](https://pageflows.com/post/desktop-web/onboarding/superhuman/): Page Flows ## Cited in - [Do product tours still work?](https://useronboarding.ai/questions#tours) - [What counts as activated?](https://useronboarding.ai/questions#activation) --- # Userflow Source: https://useronboarding.ai/companies/userflow - Category: Onboarding that uses AI - Type: Onboarding vendors - Product: https://www.userflow.com/ai Userflow's Smartflow builds an onboarding flow from a written purpose plus one recorded click-through of your app. The AI picks tooltips or modals and writes the copy for each step, and its Adoption Agent can launch those flows from a chat answer. ## Notes - Create flow offers 'Smart Flow Creator (BETA): Simply click your way through the flow' next to Start from scratch; the Purpose field 'directly impacts the quality of the flow' per the docs. - After the purpose, you enter the app URL and hit 'Start click-through'; a bottom recording bar counts steps (30-step limit) with Pause, Start over, Done and Cancel. - Generation is async: 'We are creating your new flow… we'll send you an email when the flow is ready.' - The generated 'Add team members' flow shows an intro modal with 'Start tutorial' followed by tooltips ('Click the gear icon in the left navigation to open Settings') each advancing on 'This element is clicked'. - The Adoption Agent (formerly AI Assistant) trains on website, knowledge base, Flows and Action flows, and can import Zendesk, Salesforce Knowledge or Intercom help centers even behind a login. - When a question matches a Flow, the agent 'returns a matching Flow alongside its written answer' so the user starts the guided walkthrough from chat; Action flows show as a card and run only after the user clicks. ## Patterns record-to-flow, purpose prompt, ai picks ui pattern, chat launches tour ## Screens - ![Create flow dialog: Smart Flow Creator (beta) with a Purpose field that steers the AI](https://useronboarding.ai/images/userflow-1.jpg) Create flow dialog: Smart Flow Creator (beta) with a Purpose field that steers the AI (source: https://help.userflow.com/docs/creating-smartflows-1) - ![Smart Flow Creation panel: purpose 'Guide users through inviting team members', app URL, and Start click-through](https://useronboarding.ai/images/userflow-2.jpg) Smart Flow Creation panel: purpose 'Guide users through inviting team members', app URL, and Start click-through (source: https://help.userflow.com/docs/creating-smartflows-1) - ![Annotated recording bar that counts captured steps while the builder clicks through the app](https://useronboarding.ai/images/userflow-3.jpg) Annotated recording bar that counts captured steps while the builder clicks through the app (source: https://help.userflow.com/docs/creating-smartflows-1) - ![The AI-generated 'Add team members' flow: intro modal plus click-triggered tooltips in the builder](https://useronboarding.ai/images/userflow-4.jpg) The AI-generated 'Add team members' flow: intro modal plus click-triggered tooltips in the builder (source: https://help.userflow.com/docs/creating-smartflows-1) ## Articles - [Creating AI-generated Smartflows](https://help.userflow.com/docs/creating-smartflows-1): Userflow Help Center (by the product team) - [AI Assistant setup and knowledge base configuration](https://help.userflow.com/docs/ai-assistant-1): Userflow Help Center (by the product team) - [AI Onboarding Flows to Build in 2026](https://www.userflow.com/blog/ai-onboarding-flows-saas): Userflow Blog (by the product team) ## Cited in - [How do AI products set expectations about mistakes and data?](https://useronboarding.ai/questions#trust) - [What are onboarding-tool vendors doing with AI?](https://useronboarding.ai/questions#vendors) --- # Webflow (AI site builder) Source: https://useronboarding.ai/companies/webflow-ai-site-builder - Category: Onboarding that uses AI - Type: Creative - Product: https://webflow.com/ai-site-builder Webflow's AI site builder replaces the blank canvas with a prompt. You describe the site and its audience, then shape a generated multi-page draft in three steps (structure, theme, polish) before dropping into the full Designer. ## Notes - The flow has four steps on the product page: 'Describe your vision' (share what you're building and who it's for), 'Add some structure' (add, remove or reorder sections and pages), 'Make it your own', and 'Go further' in the full Webflow editor. - The prompt box reads 'Tell us what your site is all about. Include helpful information like the audience and goal', and the page offers preset prompts (Fintech, Enterprise AI startup, Design agency). - Theme customization covers '8 editable categories, including animation presets with GSAP'; the illustration shows swappable heading/paragraph/color themes over the generated page. - The output is 'a functional site and a foundational design system' built on Flowkit, Webflow's CSS framework, not a one-off layout. - Launched February 2025; after generation, Webflow AI keeps helping inside the editor with new pages, CMS items and SEO/AEO. ## Patterns prompt-to-workspace, generated first project, structure-then-style wizard ## Screens - ![Step 1 prompt: 'Tell us what your site is all about. Include helpful information like the audience and goal.' (converted from AVIF)](https://useronboarding.ai/images/webflow-1.jpg) Step 1 prompt: 'Tell us what your site is all about. Include helpful information like the audience and goal.' (converted from AVIF) (source: https://webflow.com/ai-site-builder) - ![Step 2: adding a Hero section to the generated structure (converted from AVIF)](https://useronboarding.ai/images/webflow-2.jpg) Step 2: adding a Hero section to the generated structure (converted from AVIF) (source: https://webflow.com/ai-site-builder) - ![Step 3: theme picker restyling the generated 'Online Store' page (converted from AVIF)](https://useronboarding.ai/images/webflow-3.jpg) Step 3: theme picker restyling the generated 'Online Store' page (converted from AVIF) (source: https://webflow.com/ai-site-builder) - ![Generated page for the example floral studio prompt (converted from AVIF)](https://useronboarding.ai/images/webflow-4.jpg) Generated page for the example floral studio prompt (converted from AVIF) (source: https://webflow.com/ai-site-builder) ## Articles - [Build sites that scale with Webflow's AI site builder](https://webflow.com/blog/ai-site-builder): Webflow Blog (by the product team) - [Build a site with Webflow's AI site builder](https://help.webflow.com/hc/en-us/articles/38840145286035-Build-a-site-with-Webflow-s-AI-site-builder): Webflow Help Center (by the product team) - [Webflow Unveils AI Prompt-to-Production, a New Way to Build with AI](https://www.prnewswire.com/news-releases/webflow-unveils-ai-prompt-to-production-a-new-way-to-build-with-ai-302558866.html): PR Newswire (Webflow) (by the product team) ## Cited in - [Can you skip the questionnaire by reading context the user already has?](https://useronboarding.ai/questions#context-not-questions) - [Should the first output be finished or a draft?](https://useronboarding.ai/questions#draft-first) --- # Whatfix Source: https://useronboarding.ai/companies/whatfix - Category: Onboarding that uses AI - Type: Onboarding vendors - Product: https://whatfix.com/ai/ Whatfix, a digital adoption platform for employee software rollouts, uses AI agents on both sides. Authors prompt pop-ups, walkthroughs and display rules into existing apps, and new employees type a task like 'Help me submit a PTO request' to get pointed to the first click. ## Notes - Authoring Agent 'generates fully configured in-app experiences using simple natural language prompts', including pop-ups, walkthroughs and visibility rules (press release, Sept 9, 2025). - Product page tiles show the authoring loop: 'Generate a downtime announcement' drafts a pop-up; 'Customise with AI' restyles it from text ('Make this look like an alert pop-up'); 'Create rules using AI' turns 'show this pop-up only on gmail… between 10 AM to 11 PM' into targeting. - Guidance Agent answers in the workflow from internal docs; the tile shows 'Help me submit a PTO request' anchoring a 'Click here to get started' callout on the Request time off element. - All agents run on ScreenSense, Whatfix's context engine built on a computer-use model trained on GUI interactions, which reads what screen and step the user is on (Feb 2025 announcement). - The onboarding case studies Whatfix cites are about employees: Acorn's '80% acceleration' in recruiter onboarding and REG's '50% improvement in time to productivity.' ## Patterns prompt-to-flow builder, natural-language targeting rules, task-intent guidance, employee onboarding ## Screens - ![Authoring Agent: 'Generate a downtime announcement' drafts an in-app pop-up](https://useronboarding.ai/images/whatfix-1.jpg) Authoring Agent: 'Generate a downtime announcement' drafts an in-app pop-up (source: https://whatfix.com/ai/) - !['Customise with AI' restyles a pop-up from a text instruction](https://useronboarding.ai/images/whatfix-2.jpg) 'Customise with AI' restyles a pop-up from a text instruction (source: https://whatfix.com/ai/) - !['Create rules using AI': plain-language display rules for a pop-up](https://useronboarding.ai/images/whatfix-3.jpg) 'Create rules using AI': plain-language display rules for a pop-up (source: https://whatfix.com/ai/) - ![Guidance Agent: 'Help me submit a PTO request' points to the Request time off element](https://useronboarding.ai/images/whatfix-4.jpg) Guidance Agent: 'Help me submit a PTO request' points to the Request time off element (source: https://whatfix.com/ai/) ## Articles - [Whatfix Launches AI Agents to Accelerate Business Outcomes for Enterprises](https://whatfix.com/newsroom/press-releases/whatfix-launches-ai-agents-to-accelerate-business-outcomes-for-enterprises/): Whatfix Newsroom (by the product team) - [Whatfix Unveils ScreenSense: An AI Technology to Shape the Next Frontier of Digital Adoption](https://whatfix.com/newsroom/press-releases/whatfix-unveils-screensense-an-ai-technology-to-shape-the-next-frontier-of-digital-adoption/): Whatfix Newsroom (by the product team) - [Whatfix AI](https://whatfix.com/ai/): Whatfix product page (by the product team) ## Cited in - [What are onboarding-tool vendors doing with AI?](https://useronboarding.ai/questions#vendors) --- # Duolingo Source: https://useronboarding.ai/companies/duolingo - Category: Onboarding that uses AI - Type: Other - Product: https://www.duolingo.com Duolingo's onboarding asks for a daily goal, then lets learners who already know some of the language take an adaptive placement test. The test uses real exercises chosen one at a time by a model to decide where in the course they start. ## Notes - Onboarding splits early: 'New to Korean? Start here with the Basics' or 'Already know some Korean? Try this Placement Test' (GoodUX teardown screenshots). - The placement test uses normal lesson exercises (e.g. 'What sound does this make?') with hearts and a progress bar, so it doesn't feel like a separate exam. - Research Director Burr Settles: 'The difficulty of the words, the grammar, and the way we present it to you in the test, all play a role to pick the exact configuration so that in less than five minutes we have a really good sense of where you're going to start the course' (AWS customer story). - The same modelling keeps personalizing after onboarding: spaced repetition that predicts recall per word and adds practice at the right time (AWS); Duolingo's blog says AI lets it 'curate and recommend resources for learners that are at a really specific difficulty level.' - The goal picker (Casual 5 min to Intense 20 min/day) comes with the mascot reassuring 'You can always change goals later.' ## Patterns adaptive placement, persona/goal quiz, skip-ahead for experienced users ## Screens - !['Pick a goal': daily-minutes goal with Duo saying you can change it later](https://useronboarding.ai/images/duolingo-1.jpg) 'Pick a goal': daily-minutes goal with Duo saying you can change it later (source: https://goodux.appcues.com/blog/duolingo-user-onboarding) - !['Choose a path': start from basics or take the placement test](https://useronboarding.ai/images/duolingo-2.jpg) 'Choose a path': start from basics or take the placement test (source: https://goodux.appcues.com/blog/duolingo-user-onboarding) - ![Korean placement test question built from a normal lesson exercise](https://useronboarding.ai/images/duolingo-3.jpg) Korean placement test question built from a normal lesson exercise (source: https://goodux.appcues.com/blog/duolingo-user-onboarding) ## Articles - [How AI is helping Duolingo democratize education](https://blog.duolingo.com/ai-democratize-education/): Duolingo Blog (by the product team) - [Duolingo (AWS machine learning customer story)](https://aws.amazon.com/machine-learning/customers/innovators/duolingo): AWS - [Duolingo's delightful user onboarding experience](https://goodux.appcues.com/blog/duolingo-user-onboarding): GoodUX by Appcues ## Cited in - [What can't this data tell us?](https://useronboarding.ai/questions#gaps) --- # Frigade Source: https://useronboarding.ai/companies/frigade - Category: Onboarding that uses AI - Type: Onboarding vendors - Product: https://frigade.com/ Frigade embeds an AI assistant that learns a product by using it, then onboards end users by answering in context, highlighting the next UI element step by step, or doing the setup task for them. ## Notes - Learning is hands-off: the assistant 'takes a seat like any user' with a test login, works through real workflows, and 're-learns as you ship' on a schedule, flagging moved UI (homepage). - The homepage demo shows 'How do I see my analytics?' turned into a live walk-through: a dark hint bar ('Open the drop-down menu', then 'Tap on Analytics') with a blue marker on the exact menu item. - For 'Invite Christian@acme.com to my account' the assistant navigates to Workspace → People, pre-fills the email and says 'Here you go! Just hit the Invite button', leaving the final click to the user. - Skills (launched July 14, 2026) let the assistant run actions itself; one launch example is new users configuring a workspace by describing the setup they want. - Assist API (Sept 8, 2026) gives a company's own agent Frigade's product knowledge, so that agent stays the only one users talk to during onboarding. - Install is pitched as 15 minutes, trained on staging first; self-hosting with your own LLM keys is available. ## Patterns in-product ai guide, highlight-next-element, agent does setup task, self-updating tours ## Screens - ![Frame from the homepage demo video: the assistant answers 'How do I see my analytics?' with a 'Tap on Analytics' hint pinned to the menu item](https://useronboarding.ai/images/frigade-1.jpg) Frame from the homepage demo video: the assistant answers 'How do I see my analytics?' with a 'Tap on Analytics' hint pinned to the menu item (source: https://frigade.com/) - ![Frame from the homepage demo video: after an invite request, the assistant opens People, pre-fills the email and points to the Invite button](https://useronboarding.ai/images/frigade-2.jpg) Frame from the homepage demo video: after an invite request, the assistant opens People, pre-fills the email and points to the Invite button (source: https://frigade.com/) ## Articles - [Frigade Launches Skills, Giving Any Product an AI Assistant That Takes Action for Users, No Code Required](https://www.prnewswire.com/news-releases/frigade-launches-skills-giving-any-product-an-ai-assistant-that-takes-action-for-users-no-code-required-302824744.html): PR Newswire (Frigade) (by the product team) - [Frigade Launches Assist API, Turning a Company's AI Agent Into an Onboarding and Support Specialist](https://www.prnewswire.com/news-releases/frigade-launches-assist-api-turning-a-companys-ai-agent-into-an-onboarding-and-support-specialist-302872054.html): PR Newswire (Frigade) (by the product team) - [Your product can't explain itself](https://frigade.com/blog/your-product-cant-explain-itself): Frigade Blog (by the product team) - [Launch YC: Frigade: Product onboarding that builds itself](https://www.ycombinator.com/launches/Owy-frigade-product-onboarding-that-builds-itself): Y Combinator (by the product team) ## Cited in - [How do AI products set expectations about mistakes and data?](https://useronboarding.ai/questions#trust) - [What are onboarding-tool vendors doing with AI?](https://useronboarding.ai/questions#vendors) --- # Granola Source: https://useronboarding.ai/companies/granola - Category: Onboarding for AI products - Type: Productivity - Product: https://www.granola.ai Sign in with a work Google or Microsoft account, watch a founder video that demos a live meeting, and land on a 'Get started with Granola' note built from that demo. Calendar sync means upcoming meetings are already there. The only required permission is microphone and system audio. ## Notes - The 'Sign in to Download' page offers Google, Microsoft or SSO and asks for 'your work email, or the account you use for meetings'. The meetingnotes.com teardown says free providers like Gmail are blocked. - Onboarding centers on a founder video ('two minutes, I'm going to show you how') played inside the app. - The video produces a 'Get started with Granola' note (interface overview, note-taking process, meeting-end workflow) with a 'Replay video' button, so the demo output doubles as the tutorial. - Per the teardown, after login 'all your upcoming meetings are already there', and you could 'sign up ten minutes before a call and record it perfectly'. It also faults the lack of a checklist or nudges toward Recipes after the video. - Granola's own guide (Jo Barrow) says to set up once: connect the calendar and fill in role, company, focus areas and meeting language, which give Granola context for its notes. - Granola captures computer audio instead of adding a bot to the call, so there is no bot-invite step; the team explains the choice in 'Why Granola doesn't use a bot'. ## Patterns work-email gate, founder video, demo output as tutorial, calendar auto-sync, minimal permissions ## Screens - !['Sign in to Download' with Google, Microsoft or SSO and a work-email prompt](https://useronboarding.ai/images/granola-1.jpg) 'Sign in to Download' with Google, Microsoft or SSO and a work-email prompt (source: https://meetingnotes.com/blog/granola-ai-teardown) - ![In-app founder onboarding video](https://useronboarding.ai/images/granola-2.jpg) In-app founder onboarding video (source: https://meetingnotes.com/blog/granola-ai-teardown) - !['Get started with Granola' note generated from the demo, with Replay video](https://useronboarding.ai/images/granola-3.jpg) 'Get started with Granola' note generated from the demo, with Replay video (source: https://meetingnotes.com/blog/granola-ai-teardown) ## Articles - [How to get the best from Granola](https://www.granola.ai/blog/get-the-best-from-granola): Granola blog (by the product team) - [Why Granola doesn't use a bot](https://www.granola.ai/blog/why-granola-doesnt-use-a-bot): Granola blog (by the product team) - [Granola.ai Teardown: Great Notes, Too Much Friction To Share](https://meetingnotes.com/blog/granola-ai-teardown): meetingnotes.com ## Cited in - [Can you skip the questionnaire by reading context the user already has?](https://useronboarding.ai/questions#context-not-questions) - [Signup first, or value first?](https://useronboarding.ai/questions#signup-timing) --- # HubSpot (Breeze) Source: https://useronboarding.ai/companies/hubspot-breeze - Category: Onboarding that uses AI - Type: Productivity - Product: https://www.hubspot.com/products/artificial-intelligence HubSpot seeds its AI context from your website. You enter a URL, Breeze crawls the site and fills a Brand Identity profile (industry, competitors, content themes, tech stack) and optionally a brand voice, which every later AI feature reads from. ## Notes - Setup starts from a domain: the company name and domain in Account Defaults are required because 'the domain is used by HubSpot's AI to crawl your website' (KB: Generate your brand identity context with AI). - The 'Enter a website' dialog promises 'We won't overwrite anything you or your team entered manually', and brand voice is opt-in via a 'Yes, update my Brand Voice' checkbox. - The crawl fills six context groups: industry classification, customer sentiment, competitive landscape, content themes, tech stack and social responsibility; each row shows its source domain and a count of missing fields. - Brand voice setup generates up to 4 personality and 4 tone traits from a writing sample, then asks the user to review and edit them (Brand Voice product page). - HubSpot's own 'Set up a HubSpot account to adopt AI features' guide orders setup as: turn on AI, grant data access, crawl for brand identity, complete company profile, then define ICPs and personas. - Crawl and brand identity context are gated to Professional and Enterprise editions. ## Patterns url-to-workspace, ai-prefilled profile, review-and-edit generated defaults, manual-edits-protected ## Screens - !['Enter a website' dialog: URL field with redirect check and an opt-in checkbox to update Brand Voice from the crawl](https://useronboarding.ai/images/hubspot-breeze-1.jpg) 'Enter a website' dialog: URL field with redirect check and an opt-in checkbox to update Brand Voice from the crawl (source: https://knowledge.hubspot.com/branding/generate-your-brand-identity-context-with-ai) - ![Additional Context table filled from the crawled site, each field tagged with its source and a 'Missing' count](https://useronboarding.ai/images/hubspot-breeze-2.jpg) Additional Context table filled from the crawled site, each field tagged with its source and a 'Missing' count (source: https://knowledge.hubspot.com/branding/generate-your-brand-identity-context-with-ai) - !['Review and edit brand voice': AI-generated personality and tone chips from a writing sample](https://useronboarding.ai/images/hubspot-breeze-3.jpg) 'Review and edit brand voice': AI-generated personality and tone chips from a writing sample (source: https://www.hubspot.com/products/content/brand-voice) ## Articles - [Generate your brand identity context with AI](https://knowledge.hubspot.com/branding/generate-your-brand-identity-context-with-ai): HubSpot Knowledge Base (by the product team) - [Set up a HubSpot account to adopt AI features](https://knowledge.hubspot.com/ai/set-up-a-hubspot-account-to-adopt-ai-features): HubSpot Knowledge Base (by the product team) - [AI Brand Voice](https://www.hubspot.com/products/content/brand-voice): HubSpot product page (by the product team) ## Cited in - [Can you skip the questionnaire by reading context the user already has?](https://useronboarding.ai/questions#context-not-questions) --- # Noom Source: https://useronboarding.ai/companies/noom - Category: Onboarding that uses AI - Type: Other - Product: https://www.noom.com Noom's web onboarding is a quiz of up to 113 screens that builds a personalized weight-loss plan. It keeps redrawing a predicted goal date as the user answers and shows 'analyzing' loaders that frame the plan as computed from their profile. It is algorithmic personalization; the teardown does not describe any of it as AI. ## Notes - Up to 113 screens and 10–15 minutes before the paywall (RevenueCat teardown by Daphne Tideman, April 2026). - A projection chart ('We predict you'll be… by [date]') recurs through the quiz, and 'as users provide more answers, Noom moves dates forward and sharpens projections', which the teardown credits with motivating completion. - A 10-question behavioral profile quiz (slider statements) ends in an 'Analyzing your profile…' loader and a named profile result card. - Later loaders such as 'Analyzing demographic profile' and 'Cross-checking with user database' insert extra questions while 'processing', which makes the plan feel computed from the answers. - Caveat for this gallery: none of the cited sources call Noom's plan generation AI. It is included as the reference pattern for data-driven personalized-plan onboarding that AI-native apps copy. ## Patterns persona quiz, personalized plan, live goal projection, labor-illusion loaders ## Screens - !['Analyzing your profile…' loader, then the behavioral profile result](https://useronboarding.ai/images/noom-1.jpg) 'Analyzing your profile…' loader, then the behavioral profile result (source: https://www.revenuecat.com/blog/growth/web-to-app-onboarding-funnel) - ![Goal-date projection chart that moves as answers come in](https://useronboarding.ai/images/noom-2.jpg) Goal-date projection chart that moves as answers come in (source: https://www.revenuecat.com/blog/growth/web-to-app-onboarding-funnel) - ![Updated prediction screen with the personalized weight curve and target date](https://useronboarding.ai/images/noom-3.jpg) Updated prediction screen with the personalized weight curve and target date (source: https://www.revenuecat.com/blog/growth/web-to-app-onboarding-funnel) - !['Analyzing demographic profile' / 'Cross-checking with user database' loaders with questions mixed in](https://useronboarding.ai/images/noom-4.jpg) 'Analyzing demographic profile' / 'Cross-checking with user database' loaders with questions mixed in (source: https://www.revenuecat.com/blog/growth/web-to-app-onboarding-funnel) ## Articles - [Inside Noom's Web-to-App Onboarding Funnel: UX Teardown + Key Takeaways](https://www.revenuecat.com/blog/growth/web-to-app-onboarding-funnel): RevenueCat Blog - [The 113-screen onboarding that doesn't feel long](https://www.growthwaves.io/p/the-113-screen-onboarding-that-doesnt): Growthwaves ## Cited in - [Should you ask questions before the first output?](https://useronboarding.ai/questions#ask-or-infer) - [What can't this data tell us?](https://useronboarding.ai/questions#gaps) --- # Notion AI Source: https://useronboarding.ai/companies/notion-ai - Category: Onboarding for AI products - Type: Productivity - Product: https://www.notion.com/product/ai No separate AI onboarding. AI opens inside the page the user is editing, with starter prompts grouped by job (write, edit, summarize). Picking one seeds a focused follow-up, and the output lands in the document. ## Notes - AI UX Playground: 'There is no empty chat screen'; starters such as 'Write meeting agenda' appear in the editor where the cursor already is. - The Notion AI panel headline is 'How can I help you today?' above a 'Get started' row of one-click starters. - Starters are grouped by job (write, edit, summarize) so users 'browse by intent, not by prompt engineering skill'. - Selecting 'Write meeting agenda' opens a tighter AI surface seeded with the starter and inserts the output into the page instead of a side thread. - Critique from the teardown: first-time users may not know AI exists until they invoke it, and the selection scope (block vs page) is implicit. - Designer Ryo Lu explains that Notion AI uses context to "get you closer to the AI tools you need". On a new page, Space for AI puts drafting first (brainstorm, blog post, outline). On a page with content, it puts Continue writing, Summarize and Find action items first. On selected text, it shows edit tools with a before/after comparison (Notion Blog, 2023-09-21). - The design mixes pre-packaged prompts with free text, the way templates work, so new users get an easy start and can still "prompt AI for anything in plain language" (Notion Blog, design thinking). - CEO Ivan Zhao lists the launch decisions: Space instead of the existing / command as the entry point, an interactive writing UI instead of the forms other AI tools used, and pre-packaged prompts balanced with custom input. 'Improve writing' became the most-used action (Notion Blog, behind the scenes, 2023-11-30). ## Patterns in-context activation, categorized prompt starters, no wizard, output inserted in place, context-aware prompt presets, keyboard entry point, inline AI instead of tour ## Screens - ![Notion AI 'How can I help you today?' with Get started prompt starters](https://useronboarding.ai/images/notion-ai-1.jpg) Notion AI 'How can I help you today?' with Get started prompt starters (source: https://aiuxplayground.com/teardowns/notion/onboarding) - ![After 'Write meeting agenda': seeded follow-up that creates a Meeting agenda page](https://useronboarding.ai/images/notion-ai-2.jpg) After 'Write meeting agenda': seeded follow-up that creates a Meeting agenda page (source: https://aiuxplayground.com/teardowns/notion/onboarding) ## Articles - [The design thinking behind Notion AI](https://www.notion.com/blog/the-design-thinking-behind-notion-ai): Notion Blog (Ryo Lu, Product & Design) (by the product team) - [The crucial moments and decisions leading up to the launch of Notion AI](https://www.notion.com/blog/behind-the-scenes-notion-ai): Notion Blog (Ivan Zhao, CEO) (by the product team) - [Notion AI onboarding UX: prompt starters & first value](https://aiuxplayground.com/teardowns/notion/onboarding): AI UX Playground - [How to Design AI Onboarding & First-Value UX](https://aiuxplayground.com/guides/how-to-design-ai-onboarding/): AI UX Playground ## Cited in - [How do products solve the blank prompt box?](https://useronboarding.ai/questions#blank-box) - [Do product tours still work?](https://useronboarding.ai/questions#tours) --- # Windsurf (now Devin Desktop, Cognition) Source: https://useronboarding.ai/companies/windsurf - Category: Onboarding for AI products - Type: Dev tools - Product: https://windsurf.com A desktop IDE whose first run is a short native wizard: pick a color theme, choose keybindings, optionally import settings from VS Code or Cursor, install a terminal command, then sign in. After setup the start page opens on the agent input box. ## Notes - The docs page (docs.windsurf.com/windsurf/getting-started) now 307-redirects to docs.devin.ai/desktop/getting-started; the product was rebranded to Devin Desktop and the docs note repository URLs "still use the pre-rebrand windsurf naming, but the package is now called devin-desktop". - The wizard's single "Customize" screen bundles three decisions: color theme (Light / Dark / Other, each previewed with a React code snippet), a pre-checked "Install devin-desktop terminal command" box, and a collapsed "Import settings" disclosure. - Expanding "Import settings" reveals keybinding preferences (VS Code or Vim), so switchers can keep muscle memory without visiting settings. - Migration from competitors is also exposed as command-palette actions: "Import VS Code extensions/settings" and "Import Cursor extensions/settings" — Cursor is called out by name as a switching source. - Auth comes after customization, not before: a "Welcome to Devin" screen with Sign up (primary) / Log in and a "Having trouble? Enter your auth token manually" fallback link for when the browser round-trip fails. - Per the docs, after setup the start page presents the agent input interface, with recent projects, open folder and settings as secondary paths. ## Patterns preference setup wizard, import from competitor, defaults pre-selected, auth after setup, manual auth fallback, agent prompt as home screen ## Screens - ![Customize step with Import settings expanded: theme cards, terminal-command checkbox, and VS Code vs Vim keybinding choice](https://useronboarding.ai/images/windsurf-3.jpg) Customize step with Import settings expanded: theme cards, terminal-command checkbox, and VS Code vs Vim keybinding choice (source: https://docs.devin.ai/desktop/getting-started) - ![Customize step collapsed: Light/Dark/Other theme picker with code previews and a single Next button](https://useronboarding.ai/images/windsurf-4.jpg) Customize step collapsed: Light/Dark/Other theme picker with code previews and a single Next button (source: https://docs.devin.ai/desktop/getting-started) - !["Welcome to Devin" sign-in screen: Sign up, Log in, and a manual auth-token fallback link](https://useronboarding.ai/images/windsurf-5.jpg) "Welcome to Devin" sign-in screen: Sign up, Log in, and a manual auth-token fallback link (source: https://docs.devin.ai/desktop/getting-started) - ![Command palette: "Windsurf: Import VS Code extensions" and "Import VS Code settings"](https://useronboarding.ai/images/windsurf-1.jpg) Command palette: "Windsurf: Import VS Code extensions" and "Import VS Code settings" (source: https://docs.devin.ai/desktop/getting-started) - ![Command palette: "Windsurf: Import Cursor extensions" and "Import Cursor settings"](https://useronboarding.ai/images/windsurf-2.jpg) Command palette: "Windsurf: Import Cursor extensions" and "Import Cursor settings" (source: https://docs.devin.ai/desktop/getting-started) ## Articles - [Getting started (Devin Desktop, formerly Windsurf)](https://docs.devin.ai/desktop/getting-started): Devin Desktop docs (by the product team) ## Cited in - [Can you skip the questionnaire by reading context the user already has?](https://useronboarding.ai/questions#context-not-questions) --- # Chameleon Source: https://useronboarding.ai/companies/chameleon - Category: Onboarding that uses AI - Type: Onboarding vendors - Product: https://www.chameleon.io/agents Chameleon's Copilot turns a goal like 'launch our new feature' into a researched brief and then a multi-experience onboarding campaign. It uses Chameleon's benchmark data to push back on tours that users tend to dismiss. ## Notes - Copilot opens on 'What are you trying to do?' with goal chips (Launch a new feature, Drive adoption of a feature, More…), a free-text prompt that can reference an existing walkthrough, an attached PDF and an @-picked audience segment. - Its pipeline is shown as Researching → Strategizing → Creating: a 'Researched Topic' card (it uses web search on your product), a 'Strategy Brief' with objective and problem, then generated steps with Next/Dismiss buttons. - Output includes copy, targeting rules and goal configuration across tours, modals, banners and surveys, not just a single tour. - It is trained on Chameleon Benchmarks ('5-step tours see 78% drop-off') and is pitched as protecting users by pushing back on bad UX; CEO Pulkit Agrawal cites a 91% dismissal rate on first onboarding steps. - Runs in strategy, create and analyze modes, so it can critique existing experiences as well as build new ones. ## Patterns goal-to-campaign, ai strategy brief, benchmark-grounded pushback, prompt-to-flow builder ## Screens - ![Copilot prompt: 'What are you trying to do?' with goal chips, an attached feature PDF and an audience segment picker](https://useronboarding.ai/images/chameleon-1.jpg) Copilot prompt: 'What are you trying to do?' with goal chips, an attached feature PDF and an audience segment picker (source: https://www.chameleon.io/blog/copilot) - ![From 'Launch our new Intelligence 2.0 feature' to Researched Topic, Strategy Brief and generated tour steps](https://useronboarding.ai/images/chameleon-2.jpg) From 'Launch our new Intelligence 2.0 feature' to Researched Topic, Strategy Brief and generated tour steps (source: https://www.chameleon.io/blog/copilot) ## Articles - [Meet Copilot: The UX Agent That Protects Users From Your Best Intentions](https://www.chameleon.io/blog/copilot): Chameleon Blog (Pulkit Agrawal) (by the product team) - [AI in User Onboarding: The Complete Guide for B2B SaaS](https://www.chameleon.io/blog/ai-user-onboarding): Chameleon Blog (by the product team) - [10 Ways to Use AI in User Onboarding That You Can Ship Today](https://www.chameleon.io/blog/10-ways-to-use-ai-user-onboarding): Chameleon Blog (by the product team) ## Cited in - [Do product tours still work?](https://useronboarding.ai/questions#tours) - [What are onboarding-tool vendors doing with AI?](https://useronboarding.ai/questions#vendors) --- # Dia (The Browser Company) Source: https://useronboarding.ai/companies/dia - Category: Onboarding for AI products - Type: Productivity - Product: https://www.diabrowser.com An AI browser whose first-run 'Getting Started' page teaches five things: trust, the ⌘E chat shortcut, learning your patterns, history as context, and payoff by the end of the week. Setup offers optional import from your old browser and ChatGPT memories, with preloaded Skills. ## Notes - 'Welcome to Dia — Here are five things to know': 1 Built for your trust (data encrypted and stored locally), 2 With you on every page ('Press ⌘E to open Chat'), 3 Learns how you work, 4 History, but way better, 5 By the end of the week… - The page sets a time-to-value promise: 'After a week, you won't be able to go back to your old browser.' - Privacy is placed first; the history section says Dia avoids storing data from sensitive sites such as banking and health by default. - Illustrations show the chat bar with sample prompts ('suggest ways this website could be more accessible', 'What was that research paper… I looked at last week?') and Skill chips (Meeting Recap, Show Up Smart, Focus Brief). - Dia's help copy calls import of bookmarks, passwords, history and extensions optional and offers 'Import from ChatGPT…' to seed Memories; Skills come preloaded (e.g. Summarize in Chat). - SupaSidebar's review says first run imports from Safari, Chrome, Edge or Brave, opens the AI sidebar, and includes a short Skills walkthrough and a Memory opt-in. ## Patterns welcome page with numbered principles, keyboard shortcut teaching, privacy-first framing, import from previous tool, memory opt-in, preloaded skills ## Screens - ![Getting Started 'With you on every page': chat bar on a GitHub page with a sample prompt](https://useronboarding.ai/images/dia-1.jpg) Getting Started 'With you on every page': chat bar on a GitHub page with a sample prompt (source: https://www.diabrowser.com/getting-started) - ![Chat bar with Skill chips (Meeting Recap, Show Up Smart, Focus Brief) and a history-recall prompt](https://useronboarding.ai/images/dia-2.jpg) Chat bar with Skill chips (Meeting Recap, Show Up Smart, Focus Brief) and a history-recall prompt (source: https://www.diabrowser.com/getting-started) - !['What are my priorities this week?' answered from browsing context](https://useronboarding.ai/images/dia-3.jpg) 'What are my priorities this week?' answered from browsing context (source: https://www.diabrowser.com/getting-started) ## Articles - [Getting Started (Welcome to Dia)](https://www.diabrowser.com/getting-started): Dia Browser (by the product team) - [Dia Browser Mac Review (2026)](https://supasidebar.com/blog/dia-browser-mac-review-2026): SupaSidebar ## Cited in - [Can you skip the questionnaire by reading context the user already has?](https://useronboarding.ai/questions#context-not-questions) - [How do products teach people to prompt?](https://useronboarding.ai/questions#teach-prompting) - [How do AI products set expectations about mistakes and data?](https://useronboarding.ai/questions#trust) - [What can't this data tell us?](https://useronboarding.ai/questions#gaps) --- # Otter.ai Source: https://useronboarding.ai/companies/otter - Category: Onboarding for AI products - Type: Productivity - Product: https://otter.ai A welcome survey (work type, role, goals), a calendar step that suggests meetings for the AI notetaker to join automatically, workspace and teammate invites, and then in-product tooltips around a tutorial recording. ## Notes - UserGuiding's composite shows the survey order: 'Get automatic meeting notes' (join and share settings), 'What best describes your work?' chips, 'What's your role?', and 'How would you like Otter to help you?'. - A calendar step, 'Suggested meetings to record from your calendar', lists upcoming meetings with per-meeting Auto-join toggles. - Workspace steps: 'Boost team productivity with workspace' and 'Invite teammates to your workspace' with suggested colleagues pre-listed. - Inside the app, a 'My Recording Tutorial' note is paired with a 'Stay Aligned' tooltip ('Post the automated meeting summary directly in Slack') and an 'Ask Otter anything about your conversations' box; the Basic plan meter shows '0 of 300 monthly minutes used'. - UserGuiding rates it 4/5: checklists and meeting-based aha moments work, but flows 'lack a clear sense of progress', CTAs dilute focus, and untriggered pop-ups can disrupt. - Otter's help docs describe connecting Google or Microsoft calendars so the notetaker auto-joins Zoom, Meet and Teams meetings. ## Patterns welcome survey, calendar-based suggestions, auto-join toggles, teammate invites, tooltips, usage meter ## Screens - ![Composite of Otter signup: welcome survey, suggested meetings to auto-record, workspace and invite steps (converted from AVIF)](https://useronboarding.ai/images/otter-1.jpg) Composite of Otter signup: welcome survey, suggested meetings to auto-record, workspace and invite steps (converted from AVIF) (source: https://userguiding.com/blog/saas-onboarding-ux-analysis) - !['My Recording Tutorial' with a 'Stay Aligned' Slack tooltip (converted from AVIF)](https://useronboarding.ai/images/otter-2.jpg) 'My Recording Tutorial' with a 'Stay Aligned' Slack tooltip (converted from AVIF) (source: https://userguiding.com/blog/saas-onboarding-ux-analysis) ## Articles - [Otter Quick Start Guide](https://help.otter.ai/hc/en-us/articles/360049722894-Otter-Quick-Start-Guide): Otter Help Center (by the product team) - [Connect your Calendar and Contacts to Otter](https://help.otter.ai/hc/en-us/articles/360048070154-Connect-your-Calendar-and-Contacts-to-Otter): Otter Help Center (by the product team) - [We Analyzed 36 SaaS Onboarding Flows and Rated Every Single One](https://userguiding.com/blog/saas-onboarding-ux-analysis): UserGuiding ## Cited in - [Can you skip the questionnaire by reading context the user already has?](https://useronboarding.ai/questions#context-not-questions) --- # Pendo Source: https://useronboarding.ai/companies/pendo - Category: Onboarding that uses AI - Type: Onboarding vendors - Product: https://www.pendo.io/product/in-app-guides/ Pendo's guide editor now opens with 'Start with a prompt': a product team describes the onboarding tour and who it's for, and the Guides Agent picks a layout and drafts the steps and copy for review in a two-panel editor. ## Notes - The create screen offers 'Start with a prompt' ('Tell us what your guide should do and who it's for. We'll pick a layout and draft the first version for you') next to 'Start with a layout', where AI still writes the first draft of a Walkthrough or Announcement. - The Summer Release post pitches it as: describe 'an onboarding tour for new users' and 'the Pendo Guides Agent… drafts the layout, content, and flow.' - Follow-up prompts can edit text blocks, images and buttons or switch a draft between walkthrough and announcement formats; the AI 'doesn't access your application or visitor data.' - Beta gaps: no polls, activation settings or localizations from the AI editor; advanced styling still sends you to the Visual Design Studio. - Requires Guides Pro and AI settings enabled by a subscription admin; live preview needs web SDK 2.339.1 or later. ## Patterns prompt-to-flow builder, layout-first AI draft, draft-then-edit ## Screens - !['AI-assisted options': Start with a prompt, or pick a Walkthrough/Announcement layout for AI to fill](https://useronboarding.ai/images/pendo-1.jpg) 'AI-assisted options': Start with a prompt, or pick a Walkthrough/Announcement layout for AI to fill (source: https://support.pendo.io/hc/en-us/articles/40409074705691-Create-a-guide-using-conversational-AI-beta) - ![A generated draft guide ('Notes announcement – benefit-focused') with Edit in Pendo and Edit in my app options](https://useronboarding.ai/images/pendo-2.jpg) A generated draft guide ('Notes announcement – benefit-focused') with Edit in Pendo and Edit in my app options (source: https://support.pendo.io/hc/en-us/articles/40409074705691-Create-a-guide-using-conversational-AI-beta) ## Articles - [Create a guide using conversational AI (beta)](https://support.pendo.io/hc/en-us/articles/40409074705691-Create-a-guide-using-conversational-AI-beta): Pendo Help Center (by the product team) - [How Pendo's Summer Release reimagines onboarding, support, and expansion in the SaaS + AI era](https://www.pendo.io/pendo-blog/building-an-intelligent-foundation/): Pendo Blog (by the product team) - [Introducing Agent Mode in Pendo](https://www.pendo.io/pendo-blog/agent-mode-pendo-experience/): Pendo Blog (by the product team) ## Cited in - [What are onboarding-tool vendors doing with AI?](https://useronboarding.ai/questions#vendors) --- # Replit Source: https://useronboarding.ai/companies/replit - Category: Onboarding for AI products - Type: Dev tools - Product: https://replit.com Replit's home screen is a single Agent prompt box greeting the user by name, with suggested prompts and an artifact chip (e.g. Website). Replit's own first-app guide teaches users to write an outcome-plus-success-criteria prompt, then drops them into a split chat/preview workspace while Agent plans and builds. ## Notes - The home screen headline is personalized ("Raouf, what are we working on today?") and sits above a "Suggested for you" row with a refresh icon, including "Look me up and suggest something" and a Notion-connector prompt ("Turn our Notion meeting notes into tasks…"). - The prompt box carries an artifact-type chip ("Website"), a mode selector ("Free"), a mic button and an attach (+) button; the docs say artifact selection "makes intentions explicit" and mode selection determines resource allocation. - The official "Build your first app" guide models prompt-writing as onboarding: the example prompt "starts with one outcome… defines observable success criteria, and postpones login, data, payments, and integrations." - After submit the user lands in a two-pane workspace: chat on the left with a "Planning artifact creation" step and Agent's restated plan, live Preview on the right, with Design/Build tabs and Invite/Publish buttons already visible. - The guide sets a time expectation — about 10 minutes for the first build, with a warning that Free Mode can take several minutes for complex requests — then pushes a test-in-Preview, refine-with-one-prompt loop. - Replit's 2025 review says the December free tier came with "groundbreaking improvements to the first build experience": the first build dropped from 15-20 minutes to 3-5 minutes. In June, a new app-creation flow "let you visualize what Agent would build before committing" (Replit Blog, 2025 in Review). - The Free Mode launch post describes the agent as one that "understands when you want quick answers, and when you want to do more complex work". New users can "think through a project before starting and build in the same space" (Replit Blog, 2026-08-18). ## Patterns single prompt box home, personalized greeting, suggested prompts, artifact type picker, plan before build, live preview split view, time expectation setting ## Screens - ![Home screen: personalized "what are we working on today?" prompt, suggested prompts, Website artifact chip and Free mode selector](https://useronboarding.ai/images/replit-1.jpg) Home screen: personalized "what are we working on today?" prompt, suggested prompts, Website artifact chip and Free mode selector (source: https://docs.replit.com/getting-started/quickstarts/ask-ai) - ![First build: Agent chat with "Planning artifact creation" on the left, live preview of the generated site on the right](https://useronboarding.ai/images/replit-2.jpg) First build: Agent chat with "Planning artifact creation" on the left, live preview of the generated site on the right (source: https://docs.replit.com/getting-started/quickstarts/ask-ai) ## Articles - [Build your first app](https://docs.replit.com/getting-started/quickstarts/ask-ai): Replit Docs (by the product team) - [2025: Replit in Review](https://replit.com/blog/2025-replit-in-review): Replit Blog (Matt Palmer) (by the product team) - [Replit Introduces Free Mode to Expand What is Possible with AI](https://replit.com/blog/replit-introduces-free-mode): Replit Blog (Amjad Masad, Michele Catasta, 2026-08-18) (by the product team) ## Cited in - [Where did onboarding go in AI products?](https://useronboarding.ai/questions#where-did-onboarding-go) - [How do products solve the blank prompt box?](https://useronboarding.ai/questions#blank-box) - [Should the first output be finished or a draft?](https://useronboarding.ai/questions#draft-first) - [How do products teach people to prompt?](https://useronboarding.ai/questions#teach-prompting) --- # v0 (Vercel) Source: https://useronboarding.ai/companies/v0 - Category: Onboarding for AI products - Type: Dev tools - Product: https://v0.app v0's landing page is the product: a "What do you want to create?" prompt box with one-click example chips and a template gallery. Per a third-party teardown, a logged-out visitor can submit a prompt and see v0 start generating before being asked to sign up. ## Notes - Growthmates' teardown of six AI builders found "4/6 AI tools asked me to sign up immediately, but V0 is not interrupting the first time to value" — generation starts after the prompt with no sign-up wall, and the sign-up flow is "almost invisible, with only 3 simple steps." - The logged-out home shown in the teardown has Sign In / Sign Up top-right, a model selector ("v0 Max") in the composer, and example chips under the box (Event Calendar, Shopping Cart, Music Player, User Profile, plus a refresh icon) that fill the prompt for you. - Below the fold, "Start with a template" is filtered by Apps and Games, Landing Pages, Components, Dashboards, with a Browse all link — templates are a secondary path, not a required step. - After submit, the page becomes a chat thread under a "Drafts" breadcrumb showing a "Thinking" state while Sign In / Sign Up stay in the header. - v0's own Quickstart still tells users to "Visit v0.app and sign in" first, then describe the app in natural language (examples: "A todo app with add, edit, and delete functionality"), refine in Code or Design mode, and deploy through the Publish control in the chat header. - The docs home frames onboarding by role — e.g. designers can "Clone pages with screenshots or Figma files," engineers can "Create React components." - Vercel's official prompting guide teaches new users a three-part prompt: product surface, context of use, and constraints and taste. The reasoning: vague prompts make v0 guess, and a detailed prompt came out "19 seconds faster, 152 fewer lines, lower cost" in their test. The guide treats v0 as a collaborator whose output "depends on how clearly you communicate" (Vercel Blog, 2025-12-15). ## Patterns prompt before signup, example prompt chips, template gallery, single prompt box home, role-based docs entry, prompt-writing education ## Screens - ![Logged-out v0: "What do you want to create?" prompt with example chips and template gallery (left); generation already "Thinking" with Sign In/Sign Up still in the header (right)](https://useronboarding.ai/images/v0-1.jpg) Logged-out v0: "What do you want to create?" prompt with example chips and template gallery (left); generation already "Thinking" with Sign In/Sign Up still in the header (right) (source: https://www.growthmates.news/p/your-onboarding-should-become-invisible) ## Articles - [Quickstart](https://v0.app/docs/quickstart): v0 Docs (by the product team) - [v0 Docs](https://v0.app/docs): v0 Docs (by the product team) - [How to prompt v0](https://vercel.com/blog/how-to-prompt-v0): Vercel Blog (Esteban Suárez, 2025-12-15) (by the product team) - [Your onboarding should become invisible](https://www.growthmates.news/p/your-onboarding-should-become-invisible): Growthmates (Kate Syuma) ## Cited in - [Where did onboarding go in AI products?](https://useronboarding.ai/questions#where-did-onboarding-go) - [How do products solve the blank prompt box?](https://useronboarding.ai/questions#blank-box) - [Signup first, or value first?](https://useronboarding.ai/questions#signup-timing) - [How do products teach people to prompt?](https://useronboarding.ai/questions#teach-prompting) --- # Appcues Source: https://useronboarding.ai/companies/appcues - Category: Onboarding that uses AI - Type: Onboarding vendors - Product: https://www.appcues.com/features/appcues-ai Appcues AI lets teams build onboarding flows by chatting: describe the tour, upload a mockup or paste a page URL, and the Experience Builder agent drafts the flow in the builder for you to refine by chat or by hand. ## Notes - Three inputs start a flow: a plain-language request ('Welcome new users with a three-step tour'), an uploaded design mockup whose layout, copy and styling are read from the image, or a page URL / legacy flow to convert (docs: Create an experience with Appcues AI). - Output is always a draft; 'depending on what you ask for, Appcues AI either builds the experience directly or opens the builder for you to finish it', and it may first ask about goal, audience and target URL. - The screenshot shows the chat panel left of the builder: after the prompt it lists the three steps it created and offers next actions (tailor copy, add brand colors, convert steps to tooltips, target only new users, publish). - Five named agents: Advisor, Delivery Specialist (why a flow isn't showing), Segmentation Planner, Experience Builder (beta) and Growth Analyst. - Included on every plan; Appcues says customer data isn't used to train models and all AI features can be hidden from a workspace. - Theme styling still has to be applied manually after generation; publishing needs a Publisher or Admin role. ## Patterns prompt-to-flow builder, screenshot-to-flow, chat beside canvas, agent roster ## Screens - ![Appcues AI chat creates a 3-step 'New User Welcome Tour' and opens it in the flow builder with suggested next edits](https://useronboarding.ai/images/appcues-1.jpg) Appcues AI chat creates a 3-step 'New User Welcome Tour' and opens it in the flow builder with suggested next edits (source: https://docs.appcues.com/using-the-appcues-app/create-an-experience-with-appcues-ai) ## Articles - [Create an experience with Appcues AI](https://docs.appcues.com/using-the-appcues-app/create-an-experience-with-appcues-ai): Appcues Docs (by the product team) - [Appcues AI features](https://www.appcues.com/features/appcues-ai): Appcues product page (by the product team) - [Appcues AI Prompts for Building Better Product Experiences](https://www.appcues.com/blog/appcues-ai-prompts): Appcues Blog (by the product team) ## Cited in - [What are onboarding-tool vendors doing with AI?](https://useronboarding.ai/questions#vendors) --- # Canva Source: https://useronboarding.ai/companies/canva - Category: Onboarding that uses AI - Type: Creative - Product: https://www.canva.com/magic-design/ Canva asks one use-case question at signup and then drops new users on a home screen built around a Canva AI prompt ('What will you design today?'). The first design can be generated from a sentence instead of starting from a template. ## Notes - Signup asks a single use-case question (Personal, Work, Education or Nonprofit) that 'takes two seconds' and tailors template suggestions without a tour (Supademo teardown). - The new-user home screen defaults to a Canva AI tab next to Your designs and Templates, with the prompt already filled ('create a poster ad for Supademo…') and mode chips for Design, Image, Doc and Code. - Under the prompt, a 'what you can do with AI' row shows starter prompts by type (Image: 'A cat playing with a ball of yellow yarn', Code: 'An interactive historical timeline', Write, Video). - An onboarding tooltip on the prompt bar reads 'Let's put AI to work ✨', so the AI entry point is the guided first step. - Canva's Magic Design page pitches starting from a text prompt to generate on-brand designs 'whether you're a confident Canva user or a total beginner.' ## Patterns prompt-first home, empty-state prompts, use-case question, generated first project ## Screens - ![New-user home: 'What will you design today?' with the Canva AI prompt, mode chips and 'Let's put AI to work' tooltip](https://useronboarding.ai/images/canva-1.jpg) New-user home: 'What will you design today?' with the Canva AI prompt, mode chips and 'Let's put AI to work' tooltip (source: https://supademo.com/user-flow-examples/canva) - ![Signup use-case question that personalizes template and AI suggestions](https://useronboarding.ai/images/canva-2.jpg) Signup use-case question that personalizes template and AI suggestions (source: https://supademo.com/user-flow-examples/canva) ## Articles - [Magic Design: Free Online AI Design Tool](https://www.canva.com/magic-design/): Canva (by the product team) - [Canva onboarding flow teardown](https://supademo.com/user-flow-examples/canva): Supademo ## Cited in - [How do products solve the blank prompt box?](https://useronboarding.ai/questions#blank-box) - [Should you ask questions before the first output?](https://useronboarding.ai/questions#ask-or-infer) --- # Claude Code (Anthropic) Source: https://useronboarding.ai/companies/claude-code - Category: Onboarding for AI products - Type: Dev tools - Product: https://claude.com/product/claude-code A terminal agent whose onboarding runs as a keyboard-only wizard the first time you type `claude`: an ASCII-art welcome, a text-style (theme) picker, a login-method choice that hands off to the browser, then a boxed welcome panel with tips and an empty prompt. ## Notes - First run opens with "Welcome to Claude Code v2.x", a pixel-art mascot scene, "Let's get started." and "Choose the text style that looks best with your terminal" (Dark mode / Light mode), with the hint "To change this later, run /theme" — the first question is about legibility, not the user. - The Technically Curious guide (written for non-developer business users) walks the remaining steps: pick dark/light, choose login method "1. Claude account with subscription" which opens the browser, return to the terminal, then answer an effort question ("high effort" vs "medium effort"). - The same guide warns that signing up mid-authentication is "buggy" and tells readers to create a paid Claude account before installing — the CLI assumes an existing account. - After login the session shows a bordered panel: "Welcome back Katie!", model/plan/name/working directory on the left, and "Tips for getting started" ("Run /init to create a CLAUDE.md file…", a note about launching in the home directory) plus "Recent activity" on the right, with "? for shortcuts" under the input. - Anthropic's Quickstart lists the accepted account types (Pro/Max/Team/Enterprise, Console with a "Claude Code" workspace auto-created on first login, Bedrock/Vertex-style cloud providers, or a corporate gateway) and skips the login prompt if ANTHROPIC_API_KEY is set, asking the user to approve the key instead. - The docs' suggested first prompts are codebase-orientation questions ("what does this project do?", "explain the folder structure") before any edit, then "add a hello world function to the main file". ## Patterns terminal wizard, theme picker first, browser auth handoff, tips panel, suggested first prompts, CLAUDE.md init nudge ## Screens - ![First run in a VS Code terminal: "Welcome to Claude Code" art, "Let's get started." and the Dark/Light text-style picker](https://useronboarding.ai/images/claude-code-1.jpg) First run in a VS Code terminal: "Welcome to Claude Code" art, "Let's get started." and the Dark/Light text-style picker (source: https://technicallycurious.substack.com/p/claude-code-setup-guide-windows) - ![Post-login welcome panel with account/model info, "Tips for getting started" (/init), and a first "hello" exchange](https://useronboarding.ai/images/claude-code-2.jpg) Post-login welcome panel with account/model info, "Tips for getting started" (/init), and a first "hello" exchange (source: https://technicallycurious.substack.com/p/claude-code-setup-guide-windows) ## Articles - [Quickstart](https://code.claude.com/docs/en/quickstart): Claude Code Docs (Anthropic) (by the product team) - [Claude Code Setup Guide (Windows)](https://technicallycurious.substack.com/p/claude-code-setup-guide-windows): Technically Curious (Substack) ## Cited in - [Signup first, or value first?](https://useronboarding.ai/questions#signup-timing) - [How do products teach people to prompt?](https://useronboarding.ai/questions#teach-prompting) --- # Framer Source: https://useronboarding.ai/companies/framer - Category: Onboarding that uses AI - Type: Creative - Product: https://www.framer.com/ai/ Framer's first-site experience can start from a prompt. Wireframer (May 2025) generates a responsive, structure-only layout, and the newer Agent panel builds and edits editable pages, copy and visuals on the canvas through chat. ## Notes - Wireframer 'generates clear, purpose-built layouts that focus on structure, not style', with breakpoints and placeholder content, so a new user starts from a wireframe instead of a blank canvas (Framer update, May 21, 2025). - Framer AI page: 'Describe the website, page, or update you need. Framer's AI agent creates editable pages, sections, copy, and visuals directly in your project.' - The Agent panel screenshot shows a chat beside the canvas: the request 'Change the hero to a vertical stack layout, then replace the hero image with the image attached', visible 'Thought 12s', 'Edited 7 layers, removed 2', and a Done summary. - Output is real layers (layout, typography, components, breakpoints), not an image, and users can pick the model per task. ## Patterns prompt-to-workspace, chat beside canvas, structure-first generation, generated first project ## Screens - ![Framer Agent panel editing a generated 'Bloom' site: request, thinking time, layer edits and summary](https://useronboarding.ai/images/framer-1.jpg) Framer Agent panel editing a generated 'Bloom' site: request, thinking time, layer edits and summary (source: https://www.framer.com/ai/) ## Articles - [Wireframer](https://www.framer.com/updates/wireframer): Framer Updates (by the product team) - [How to use Wireframer](https://www.framer.com/help/articles/how-to-use-wireframer/): Framer Help (by the product team) - [Framer AI](https://www.framer.com/ai/): Framer product page (by the product team) ## Cited in - [Should the first output be finished or a draft?](https://useronboarding.ai/questions#draft-first) --- # Gamma Source: https://useronboarding.ai/companies/gamma - Category: Onboarding for AI products - Type: Creative - Product: https://gamma.app Gamma's creation flow starts with a three-way choice — Paste in text, Generate, or Import file or URL — then a Generate screen where users pick a format, card count and language before typing a one-line topic. The AI returns an editable outline first, not a finished deck. ## Notes - "Create new (AI)" on the dashboard opens a full-screen "Create with AI — How would you like to get started?" chooser with three illustrated cards; Generate carries a "POPULAR" badge and the promise "Create from a one-line prompt in a few seconds." - Paste ("Create from notes, an outline, or existing content") and Import ("Enhance existing docs, presentations, or webpages") let users bring existing material instead of starting from a blank prompt — the help article pitches Paste as best for text "from notes, ChatGPT, or other sources." - The Generate screen asks "What would you like to create today?" with format tiles (Presentation, Webpage, Document, Social), an "8 cards" stepper, a layout dropdown and a language dropdown above the prompt field; a grid of "Example prompts" sits below the empty field. - The primary button is "Generate outline", so the first output is an outline the user can edit before any slides are designed — a checkpoint before spending the full generation. - The chooser also lists "Your recent prompts" with DRAFT labels, so abandoned generations are recoverable on return. - After creation, a sparkle icon on each slide opens AI editing; the help article notes multi-slide AI edits are Pro-only, and "View Source Prompt" in the … menu shows the original prompt. - Co-founder Grant Lee says Gamma spent three to four months after its launch rebuilding onboarding "so that AI was part of the actual onboarding", so every new user would feel "magic in the first 30 seconds". Signups then went from a few hundred a day to 20,000 a day (Lenny's Podcast, 2025-11-13, ~11:22). - Lee's "one egg" rule for the first run: throw a new user one idea ("create a slide in seconds"), not four or five features. He tells the team to treat new users as "selfish, vain, and lazy", with no wish to learn a new tool, and to earn each next 30 seconds (Lenny's Podcast, ~18:18-20:12). - The team's mantra for the first 30 seconds was to make it "dead simple" to create and to share. They found problems by watching usability tests of the onboarding flow and saw people "stuck on the first screen" (Lenny's Podcast, ~15:09 and ~1:06:36). ## Patterns start-mode chooser, import existing content, example prompts, outline before generation, format picker, recent drafts, AI generation as the first-run experience, single hero action ## Screens - ![Still frame from help GIF: "Create with AI — How would you like to get started?" with Paste in text, Generate (POPULAR) and Import file or URL cards](https://useronboarding.ai/images/gamma-1.jpg) Still frame from help GIF: "Create with AI — How would you like to get started?" with Paste in text, Generate (POPULAR) and Import file or URL cards (source: https://help.gamma.app/en/articles/7838093-how-do-i-create-a-new-presentation-document-or-webpage-in-gamma) - ![Still frame from help GIF: Generate screen with format tiles, 8-card stepper, language picker, prompt field and "Generate outline" button](https://useronboarding.ai/images/gamma-2.jpg) Still frame from help GIF: Generate screen with format tiles, 8-card stepper, language picker, prompt field and "Generate outline" button (source: https://help.gamma.app/en/articles/7838093-how-do-i-create-a-new-presentation-document-or-webpage-in-gamma) ## Articles - [How do I create a new presentation, document, or webpage in Gamma?](https://help.gamma.app/en/articles/7838093-how-do-i-create-a-new-presentation-document-or-webpage-in-gamma): Gamma Help Center (by the product team) - ["Dumbest idea I've heard" to $100M ARR: Inside the rise of Gamma | Grant Lee (co-founder)](https://www.lennysnewsletter.com/p/how-50-people-built-a-profitable-ai-unicorn): Lenny's Podcast (episode page; written takeaways paywalled, full episode free on YouTube: https://www.youtube.com/watch?v=3H0ngGU5pbM) (by the product team) ## Cited in - [Where did onboarding go in AI products?](https://useronboarding.ai/questions#where-did-onboarding-go) - [Should the first output be finished or a draft?](https://useronboarding.ai/questions#draft-first) - [What counts as activated?](https://useronboarding.ai/questions#activation) --- # GitHub Copilot (in VS Code) Source: https://useronboarding.ai/companies/github-copilot - Category: Onboarding for AI products - Type: Dev tools - Product: https://github.com/features/copilot Copilot onboards inside an editor the developer already uses: VS Code ships with a Copilot icon in the status bar, and signing in with GitHub from there silently enrolls anyone without a subscription in Copilot Free. There is no separate signup, trial or credit card step. ## Notes - VS Code docs: hover the Copilot icon in the Status Bar and select "Use AI Features", then "choose a sign-in method and follow the prompts"; "If you don't have a Copilot subscription yet, you'll be signed up for the Copilot Free plan." - The signed-out status-bar popover already shows working settings (Workspace Index: Locally indexed; Code Completions and Next Edit Suggestions checked) above a single "Sign in to use Copilot" button — features are presented as on, waiting only for auth. - The Accounts menu carries a badge count and a "Sign in with GitHub to use GitHub Copilot (1)" item, so the prompt is repeated in a second standard location. - GitHub's Dec 18, 2024 launch post framed the free tier as zero-friction: "2,000 code completions and 50 chat messages per month, simply by signing in with your personal GitHub account" — no trial or credit card. - After sign-in the docs route newcomers to an "agents quickstart", and usage is tracked in a Copilot status dashboard showing "the percentage of your monthly allowance that you used." ## Patterns onboard inside host tool, auto-enroll free tier, status bar entry point, existing account SSO, no credit card, usage meter ## Screens - ![Signed-out Copilot status-bar popover: pre-checked completion settings and a "Sign in to use Copilot" button](https://useronboarding.ai/images/github-copilot-1.jpg) Signed-out Copilot status-bar popover: pre-checked completion settings and a "Sign in to use Copilot" button (source: https://code.visualstudio.com/docs/copilot/setup) - ![VS Code Accounts menu with badge and "Sign in with GitHub to use GitHub Copilot (1)"](https://useronboarding.ai/images/github-copilot-2.jpg) VS Code Accounts menu with badge and "Sign in with GitHub to use GitHub Copilot (1)" (source: https://code.visualstudio.com/docs/copilot/setup) ## Articles - [Set up GitHub Copilot in VS Code](https://code.visualstudio.com/docs/copilot/setup): Visual Studio Code docs (Microsoft) (by the product team) - [Announcing 150M developers and a new free tier for GitHub Copilot in VS Code](https://github.blog/news-insights/product-news/github-copilot-in-vscode-free/): GitHub Blog (by the product team) ## Cited in - [Can you skip the questionnaire by reading context the user already has?](https://useronboarding.ai/questions#context-not-questions) - [Signup first, or value first?](https://useronboarding.ai/questions#signup-timing) --- # NotebookLM (Google) Source: https://useronboarding.ai/companies/notebooklm - Category: Onboarding for AI products - Type: Productivity - Product: https://notebooklm.google.com NotebookLM starts with the user's own sources, not a blank chat. The Add source button is the largest control in the left panel. Each upload gets an auto-generated Source Guide (summary plus key topics), and the flow runs Inputs → Chat → Outputs (Studio). ## Notes - Design lead Jason Spielman built the mental model around the creation journey "Inputs → Chat → Outputs". The Source panel is "the starting point of the user's journey", and Add source is "one click away, visibly anchored as the largest button on the panel" (jasonspielman.com/notebooklm). - Spielman kept chat as the center on purpose: "Chat remains the most intuitive and digestible way for users to engage with AI, acting as a familiar anchor", while context-aware actions surface newer features like Audio Overviews "at the right moment" (jasonspielman.com/notebooklm). - The case study says the three-panel layout took "what felt like a thousand iterations", and the team's principle is "building products with our users, not just for them" (jasonspielman.com/notebooklm). ## Patterns bring-your-own-content first run, auto summary on upload, chat as familiar anchor ## Screens - ![Annotated Sources panel: 'Add source' as the primary one-click action, with each source listed by type icon](https://useronboarding.ai/images/notebooklm-1.jpg) Annotated Sources panel: 'Add source' as the primary one-click action, with each source listed by type icon (source: https://jasonspielman.com/notebooklm) - ![Source panel states (closed, default, open), with the auto-generated Source Guide summary and Key Topics chips](https://useronboarding.ai/images/notebooklm-2.jpg) Source panel states (closed, default, open), with the auto-generated Source Guide summary and Key Topics chips (source: https://jasonspielman.com/notebooklm) ## Articles - [Designing NotebookLM](https://jasonspielman.com/notebooklm): Jason Spielman portfolio case study (by the product team) - [Google NotebookLM's Raiza Martin and Jason Spielman on Creating Delightful AI Podcast Hosts](https://sequoiacap.com/podcast/training-data-google-notebooklm): Sequoia Training Data podcast (by the product team) --- # Shopify (AI Store Builder) Source: https://useronboarding.ai/companies/shopify-ai-store-builder - Category: Onboarding that uses AI - Type: Other - Product: https://www.shopify.com/tools/ai-store-builder New merchants describe their business in a few words and Shopify generates several complete store designs, with layout, images and copy, to pick from before touching the theme editor. ## Notes - Entry point is a single 'Describe your business' prompt; the Summer '25 Editions demo types 'stylish athleisure and apparel for women' and five different-looking stores are generated (Editions image alt text). - Editions frames it as 'Design your store in seconds—describe your brand and generate a personalized theme.' - The tool page promises unlimited free generations ('generate as many store designs as you want for free') and 'no design skills needed'. - Launch coverage (May 21, 2025) describes three generated layouts with images and text per keyword prompt, replacing the pick-a-theme step of setup. ## Patterns prompt-to-workspace, multiple generated options, generated first project ## Screens - ![The AI store builder's single 'Describe your business' input (poster frame of the Editions demo video)](https://useronboarding.ai/images/shopify-ai-store-builder-1.jpg) The AI store builder's single 'Describe your business' input (poster frame of the Editions demo video) (source: https://www.shopify.com/editions/summer2025) ## Articles - [Shopify Editions Summer '25](https://www.shopify.com/editions/summer2025): Shopify (by the product team) - [Shopify AI Store Builder](https://www.shopify.com/tools/ai-store-builder): Shopify product page (by the product team) - [Shopify Debuts AI-Powered Store Builder for Merchants](https://www.pymnts.com/news/artificial-intelligence/2025/shopify-debuts-ai-powered-store-builder-for-merchants/): PYMNTS ## Cited in - [Can you skip the questionnaire by reading context the user already has?](https://useronboarding.ai/questions#context-not-questions) - [Should the first output be finished or a draft?](https://useronboarding.ai/questions#draft-first) --- # Tandem Source: https://useronboarding.ai/companies/tandem - Category: Onboarding that uses AI - Type: Onboarding vendors - Product: https://usetandem.ai/ Tandem started as an embedded AI copilot for in-app onboarding and now sells an AI workspace for B2B implementation teams. Its agents turn customer calls into plan updates that a human approves, and keep a no-login customer view of go-live progress. ## Notes - Built for 'implementation, onboarding, professional services, solutions, customer success, and forward-deployed engineering teams running complex go-lives' (homepage). - The review screen reads 'Tandem suggests: create 2 tasks & update 4 from today's call… Nothing is applied yet'; each change has a 'Why this' card quoting the call line and linking to Gong, with Apply all / Dismiss. - The customer view is a shareable link ('One link, no login, no IT ticket') showing milestones (Kickoff → Configuration → Testing & training → Go-live), 'Your next steps' synced to the customer's Jira, and hours and tasks burned. - Side panel drafts a Slack weekly report and a follow-up email to the customer contact, each with Post/Send and Edit. - 'Every agent action is approval-controlled and recorded in a complete audit log.' - Earlier positioning (still in search results for usetandem.ai/features and /enterprise) was an in-app AI agent installed with one JavaScript snippet as a side panel; fetched today, those URLs serve the implementation product instead. ## Patterns ai-drafted onboarding plan, approve-before-apply, shared customer onboarding portal, call-to-task extraction ## Screens - !['Tandem suggests: create 2 tasks & update 4 from today's call' with a 'Why this' quote from the call](https://useronboarding.ai/images/tandem-1.jpg) 'Tandem suggests: create 2 tasks & update 4 from today's call' with a 'Why this' quote from the call (source: https://usetandem.ai/) - ![Customer view of a payroll rollout: milestones, next steps synced to Jira, and AI-drafted report and follow-up](https://useronboarding.ai/images/tandem-2.jpg) Customer view of a payroll rollout: milestones, next steps synced to Jira, and AI-drafted report and follow-up (source: https://usetandem.ai/) ## Articles - [Tandem homepage](https://usetandem.ai/): Tandem (by the product team) - [Tandem vs. Intercom Fin: AI Copilot for activation](https://usetandem.ai/blog/tandem-vs-intercom-fin-ai-copilot): Tandem Blog (by the product team) ## Cited in - [How do AI products set expectations about mistakes and data?](https://useronboarding.ai/questions#trust) - [What are onboarding-tool vendors doing with AI?](https://useronboarding.ai/questions#vendors) --- # Userpilot Source: https://useronboarding.ai/companies/userpilot - Category: Onboarding that uses AI - Type: Onboarding vendors - Product: https://userpilot.com/ai/ Userpilot's agent, Lia, sits in the main nav and builds onboarding flows from a goal or from an analytics finding. Ask it to analyze activation, and it recommends a fix and offers a one-click 'Create Onboarding Flow'. ## Notes - The Lia screenshot shows a prompt-built 'New User Onboarding' flow (Step 1 welcome modal with 'Create Invoice', Step 2 tooltip 'Click here to create your first invoice') with Publish Flow / Edit Flow buttons under it. - The AI page shows analytics feeding the flow: 'Analyze my activation' returns a Seen → Clicked → Created → Published funnel, the recommendation 'Prompt users to create an invoice after sign up', and a 'Create Onboarding Flow' button. - Userpilot's blog says Lia generates 'in-app flows, checklists, and surveys around the goals you set instead of leaving you with a blank canvas.' - A separate copy assistant inside any modal, tooltip or banner writes, shortens and reframes text, produces A/B variants and translates onboarding content without an export step. - Userpilot's own guidance lists where not to use AI in onboarding: enterprise relationships, escalations in the first days, steep-learning-curve products and regulated high-liability moments. - Userpilot also ships an MCP server so Lia's product data can be queried from other AI tools. ## Patterns prompt-to-flow builder, insight-to-flow, agent in nav, ai copy variants, ai localization ## Screens - ![Lia builds a two-step 'New User Onboarding' flow from a prompt, ready to publish or edit](https://useronboarding.ai/images/userpilot-1.jpg) Lia builds a two-step 'New User Onboarding' flow from a prompt, ready to publish or edit (source: https://userpilot.com/blog/ai-user-onboarding/) - !['Analyze my activation' returns a funnel and a recommendation with a 'Create Onboarding Flow' button](https://useronboarding.ai/images/userpilot-2.jpg) 'Analyze my activation' returns a funnel and a recommendation with a 'Create Onboarding Flow' button (source: https://userpilot.com/ai/) ## Articles - [AI User Onboarding: How To Use AI Across Your Onboarding Flow (And Where Not To)](https://userpilot.com/blog/ai-user-onboarding/): Userpilot Blog (Abrar Abutouq) (by the product team) - [Userpilot AI](https://userpilot.com/ai/): Userpilot product page (by the product team) ## Cited in - [What are onboarding-tool vendors doing with AI?](https://useronboarding.ai/questions#vendors) --- # Devin (Cognition) Source: https://useronboarding.ai/companies/devin - Category: Onboarding for AI products - Type: Dev tools - Product: https://devin.ai Devin's onboarding puts repository setup before the first task. Users index a repo and configure Devin's environment, then start from a 'What do you want to get done?' box with a repo selector. The team tells new users to treat Devin like a new junior engineer. ## Notes - CEO Scott Wu says early users "see the blank screen and not really know" what to do, or asked for a whole-codebase re-architecture. Cognition changed the onboarding to steer people to "let's work on getting the repository set up first" and to give Devin "a couple one pointer tasks so it can get familiar with the code base" (Lenny's Podcast, 2025-05-04, ~17:44). - Wu's one-line onboarding framing: "treat Devin like your new junior engineer". Start with this week's easier tickets, learn what Devin needs to test its own code, then scale up (Lenny's Podcast, ~1:09:31). - Devin's docs make indexing a repository and configuring the environment prerequisites before the first session (docs.devin.ai, first run). ## Patterns setup before first task, repo selector on home, mental-model framing (junior engineer) ## Screens - ![Devin home: 'What do you want to get done?' prompt with the repository selector open ('31/155 repos set up', Connect GitHub)](https://useronboarding.ai/images/devin-1.jpg) Devin home: 'What do you want to get done?' prompt with the repository selector open ('31/155 repos set up', Connect GitHub) (source: https://docs.devin.ai/get-started/first-run) ## Articles - [Inside Devin: The AI engineer that's set to write 50% of its company's code this year | Scott Wu](https://www.youtube.com/watch?v=gI0ZNhA0rvE): Lenny's Podcast (YouTube) (by the product team) - [Devin Docs: first run](https://docs.devin.ai/get-started/first-run): Devin Docs (by the product team) ## Cited in - [How do products solve the blank prompt box?](https://useronboarding.ai/questions#blank-box) - [How do products teach people to prompt?](https://useronboarding.ai/questions#teach-prompting) - [What counts as activated?](https://useronboarding.ai/questions#activation) --- # Figma (First Draft / Figma Make) Source: https://useronboarding.ai/companies/figma - Category: Onboarding for AI products - Type: Creative - Product: https://www.figma.com Figma's AI features start from a prompt on the canvas instead of an empty frame. First Draft asks the user to pick one of four component libraries (from lo-fi wireframe to hi-fi site) and then builds an editable first mockup. ## Notes - VP of Design Noah Levin says First Draft was renamed from Make Designs to present AI output as a starting point, not a finished design. It addresses the "tedium" at the start of a project by giving designers an AI-generated first draft to iterate on (Figma Blog, 2024-09-24). - First Draft first asks for a library (App wireframe, Site wireframe, Basic app, Basic site), so the output uses real, editable components (Figma Blog). - Dylan Field at Config: "you go into Figma Design, you see a blank canvas, how do we get you to create something as fast as possible? If you go into Figma Make, how do we get you to prompt and have an awesome experience very quickly?" He also mentioned an internal team called "Blockers" whose fixes showed up as gains in activation (Lenny's Podcast live at Config, 2024-06-30). ## Patterns prompt instead of blank canvas, library picker before generation, editable first draft ## Screens - ![First Draft's first step: a 'Choose a library' picker (App wireframe, Site wireframe, Basic app, Basic site) opened from the AI button in the toolbar](https://useronboarding.ai/images/figma-1.jpg) First Draft's first step: a 'Choose a library' picker (App wireframe, Site wireframe, Basic app, Basic site) opened from the AI button in the toolbar (source: https://www.figma.com/blog/figma-ai-first-draft/) ## Articles - [Building a better First Draft for designers](https://www.figma.com/blog/figma-ai-first-draft/): Figma Blog (Noah Levin, VP of Design) (by the product team) - [Dylan Field live at Figma's Config: Intuition, simplicity, and the future of design](https://www.youtube.com/watch?v=yJQYBjto41s): Lenny's Podcast (YouTube) (by the product team) ## Cited in - [Should the first output be finished or a draft?](https://useronboarding.ai/questions#draft-first) - [What counts as activated?](https://useronboarding.ai/questions#activation) --- # Intercom (Fin) Source: https://useronboarding.ai/companies/intercom-fin - Category: Onboarding that uses AI - Type: Onboarding vendors - Product: https://www.intercom.com/fin Intercom is positioning its Fin AI agent as a 'Customer Agent' with a Success role that onboards customers after purchase: it walks them through setup step by step and suggests features they have not turned on yet, based on what they have already configured. ## Notes - Fin for Success (marked 'coming soon' in the help article) lists 'Guided onboarding: walks customers step-by-step through setup or first use', plus proactive engagement that 'spots patterns, prevents issues'. - The example screenshot shows Fin replying to 'we're only using half of what ProjectMap can do' by acknowledging completed setup ('You've already set up shared project spaces and invited your teammates') and naming two features not yet enabled, then pitching a Pro upgrade. - Fin is one agent with roles (Service, Sales, Ecommerce, Success); roles switch by conversation context and deployment audience rules rather than separate bots. - Fin for Ecommerce skips manual training: connecting Shopify syncs catalog, orders and APIs 'in a matter of minutes'. ## Patterns ai support agent in onboarding, state-aware next-step suggestions, chat-led setup ## Screens - ![Fin in a Success role: recognizes what the account has set up and recommends two unused features](https://useronboarding.ai/images/intercom-fin-1.jpg) Fin in a Success role: recognizes what the account has set up and recommends two unused features (source: https://www.intercom.com/help/en/articles/12508017-fin-as-a-customer-agent-for-service-sales-and-more) ## Articles - [Fin as a Customer Agent for service, sales, and more](https://www.intercom.com/help/en/articles/12508017-fin-as-a-customer-agent-for-service-sales-and-more): Intercom Help Center (by the product team) - [Preparing for the Customer Agent future](https://www.intercom.com/blog/preparing-for-the-customer-agent-future/): Intercom Blog (by the product team) ## Cited in - [What are onboarding-tool vendors doing with AI?](https://useronboarding.ai/questions#vendors) --- # Meta AI (standalone app) Source: https://useronboarding.ai/companies/meta-ai - Category: Onboarding for AI products - Type: Assistants - Product: https://www.meta.ai The standalone Meta AI app, launched in April 2025, is built around voice. Its home screen replaces an empty chat with an animated orb and three spoken starter options under 'Where should we start?'. Sign-in uses existing Facebook or Instagram accounts, and a Discover feed shows how other people prompt. ## Notes - Meta's launch image shows the home screen with 'Where should we start?' and three voice starters: 'Let's get to know each other', 'Learn how to use AI', 'Talk about my goals', above a 'Talk or type…' bar with a live Stop control. - Meta says the app 'gets to know your preferences, remembers context', and can draw on Facebook and Instagram if they are linked in Accounts Center, so personalization comes from existing profiles rather than an onboarding quiz. - A Discover feed lets users 'share and explore how others are using AI', which works as a prompt-inspiration surface. - Decrypt's hands-on: sign in with a Facebook or Instagram account (or an email tied to other Meta accounts); the landing page offers tutorials on using AI and starting simple projects. - The app doubles as the companion app for Meta's AI glasses (the glasses battery indicator shows in the header). ## Patterns voice-first starters, get-to-know-you prompt, existing account sign-in, social feed as inspiration ## Screens - ![Home screen: 'Where should we start?' with three voice starters](https://useronboarding.ai/images/meta-ai-1.jpg) Home screen: 'Where should we start?' with three voice starters (source: https://about.fb.com/news/2025/04/introducing-meta-ai-app-new-way-access-ai-assistant/) ## Articles - [Introducing the Meta AI App: A New Way to Access Your AI Assistant](https://about.fb.com/news/2025/04/introducing-meta-ai-app-new-way-access-ai-assistant/): Meta Newsroom (by the product team) - [Here's a Closer Look at Meta's New Standalone AI App](https://decrypt.co/316997/metas-latest-standalone-ai-app): Decrypt ## Cited in - [How do products solve the blank prompt box?](https://useronboarding.ai/questions#blank-box) - [Can you skip the questionnaire by reading context the user already has?](https://useronboarding.ai/questions#context-not-questions) - [Signup first, or value first?](https://useronboarding.ai/questions#signup-timing) - [How do voice-first products onboard?](https://useronboarding.ai/questions#voice) --- # Notion (Notion Agent) Source: https://useronboarding.ai/companies/notion - Category: Onboarding that uses AI - Type: Productivity - Product: https://www.notion.com/product/ai/use-cases/onboard-a-new-hire Notion pitches its Agent as the way to onboard a new teammate. One prompt builds a personalized onboarding page from workspace context: a welcome, week-one priorities, a database of people to meet, a tools checklist and key links. ## Notes - Notion's example prompt: 'Build an onboarding guide for a new engineer who is joining as a software engineer on the mobile team. Include a database of people to meet, a checklist of tools to get set up with, and links to important team pages.' - The agent 'uses your workspace context to aggregate the relevant teammates, tools, and pages', including connected apps such as Slack and email. - The output is ordinary Notion blocks (the example is 'Ada Onboarding — Mobile Engineer' with a welcome callout, 'Week 1 priorities' and a 'People to Meet' database) that the agent can keep editing. - Aimed at 'hiring managers and fast-growing teams'; this is AI building onboarding for the customer's own employees, not Notion's own signup flow. ## Patterns ai-generated onboarding doc, workspace-context personalization, generated first project ## Screens - ![Agent-built 'Ada Onboarding — Mobile Engineer' page with week-1 priorities and a People to Meet database](https://useronboarding.ai/images/notion-1.jpg) Agent-built 'Ada Onboarding — Mobile Engineer' page with week-1 priorities and a People to Meet database (source: https://www.notion.com/product/ai/use-cases/onboard-a-new-hire) ## Articles - [Onboard a new hire (Notion AI use case)](https://www.notion.com/product/ai/use-cases/onboard-a-new-hire): Notion (by the product team) - [How Notion Crafts a Personalized Onboarding Experience: 6 Lessons to Guide New Users](https://www.candu.ai/blog/how-notion-crafts-a-personalized-onboarding-experience-6-lessons-to-guide-new-users): Candu Blog --- # Magic Patterns Source: https://useronboarding.ai/companies/magic-patterns - Category: Onboarding for AI products - Type: Dev tools - Product: https://www.magicpatterns.com An AI prototyping tool that teaches features during the first generation instead of before it: right after sign-up, a feature-introduction modal appears over the workspace while the first output is already rendering, and the loading canvas rotates tips. ## Notes - Growthmates' teardown: "Immediately after sign-up, a feature introduction (comments, infinite canvas) appears while the first generated output is already visible." - The captured modal is a paged carousel titled "Introducing Commands" — "Begin your message with '/' to access dedicated functions that the AI is optimized for" — with back/Continue buttons, laid over the generated site and file-creation log. - While the AI generates, the preview pane shows an illustration with rotating tips (captured: "Our Github integration is a two-way sync."); the teardown says tips cover credits, shortcuts and design systems, "keeping the user engaged." - Suggested next actions sit under the first response (/Polish, Add interactivity, Select element, Visit docs), and a credit nudge appears under the composer ("Almost at your limit! Get more credits."). ## Patterns tips during generation, feature tour after first output, slash commands intro, suggested next actions, credit limit nudge ## Screens - !["Introducing Commands" modal over the first generated site (left); loading state with a GitHub-sync tip while the agent plans (right)](https://useronboarding.ai/images/magic-patterns-1.jpg) "Introducing Commands" modal over the first generated site (left); loading state with a GitHub-sync tip while the agent plans (right) (source: https://www.growthmates.news/p/your-onboarding-should-become-invisible) ## Articles - [Your onboarding should become invisible](https://www.growthmates.news/p/your-onboarding-should-become-invisible): Growthmates (Kate Syuma) ## Cited in - [Where does pricing show up?](https://useronboarding.ai/questions#pricing) - [Do product tours still work?](https://useronboarding.ai/questions#tours)