Just-in-Time Tutorials

Just-in-time tutorials reveal explanations for unfamiliar controls at the exact moment the user hovers clicks or summons them. The product stays silent until that signal arrives then drops a concise tooltip or popover loaded with the precise information needed right then. This is progressive disclosure in pure form. The old onboarding model assumed users needed a full map before they took a single step. Just in time bets that users learn better when they are already walking and the product simply hands them the next signpost. Superhuman showed the power of this in 2022. Their app detected when you reached for the mouse on actions that had keyboard shortcuts. The tooltip appeared once explained the faster path showed the actual key and never bothered you again. Loom did the same for its recording flow in 2023. Click record for the first time and you get a crisp explanation of microphone sources stop commands and trim tools. No explanation appears if you never record. Figma took it to the design canvas in 2024 and 2025. Select an auto layout frame and hover the constraint pin. A targeted explanation appears with an animated example of how the pin locks distances during resize plus a one click option to see it live on the canvas. Claude added similar hints in 2025 for its artifact toolbar. Hover the edit button on a generated React component and it explains the side by side diff mode without ever interrupting the initial generation flow. The user already engaged the control. Their brain is primed. The lesson lands and sticks because it answers a question they just asked with their cursor.

This is not the bloated product tour that paints coach marks on every button before the user has clicked anything. It is not the welcome modal that quizzes you on goals you have not formed yet. It is not the checklist that locks real work behind arbitrary progress bars or the persistent help panel that eats screen space forever. Those patterns treat the user like an empty vessel that needs filling before they are allowed to touch the product. Just in time treats the user like a competent explorer who will ask for directions by reaching for the control. Teams that cheat by triggering these tutorials on page load instead of on hover ruin the pattern and train users to ignore all future help. The irreverent truth is that most so called just in time systems shipped in early 2025 AI products were just old tours with shorter text and friendlier illustrations. Real ones wait for the users own action every single time and kill the explanation forever after one view.

Cursor delivered a masterclass with its 2025 composer update. New developers land directly in a real codebase with a blinking cursor and zero modals. When they first invoke the agent on a file with lint errors a popover explains the safety rails diff approval process and project context awareness using the exact code from their repo as the example. They dismiss it with escape and keep typing. Later while exploring the sidebar they hover the new multi file edit icon. The just in time tutorial explains it will scan the entire folder for context lists two example prompts that work best and shows a before after diff preview. When they open the terminal tab for the first time and type a natural language request the explanation appears for how the system translates to bash with automatic rollback on errors. Each explanation was scoped to a real action the developer had already chosen to take. No generic training. No video. No checklist. The result was developers shipping production changes inside their first hour. Previous versions required a separate onboarding doc that nobody read. The just in time approach killed the doc improved first session output by forty percent and let the product teach itself through real work instead of fake demos.

Reach for just in time tutorials after you have nailed direct to canvas agentic first runs smart defaults and in product nudges. They complete the system in creative tools where users probe deep into features at their own pace. Figma uses them for every new prototype interaction and variable mode control. Notion triggers database formula explanations the first time you create a rollup or button action. Midjourney explains parameter effects like stylize chaos and weird the first time a designer adjusts those sliders in 2026. AI coding tools like Cursor Lovable and v0 use them to explain agent behaviors iteration history and version branching without slowing the prompt flow. The pattern rewards curiosity and punishes nothing. Power users who know everything see zero interruptions. New users get surgical help exactly when their own exploration surfaces a question. This is how you scale complex AI products without hiring an army of support staff or producing hours of video content nobody watches. The best teams ship three or more of the five no onboarding patterns together so the just in time layer only handles the final ten percent of edge cases.

Save the just in time pattern for reversible exploratory actions inside the main product surface. Never use it to explain compliance KYC legal agreements or banking flows. Those demand explicit sequential wizards with clear checkpoints because the stakes are permanent. The same goes for SSO setup multi tenant workspace creation billing changes and any destructive action like purging a production database. Those require the user to read and acknowledge consequences before the button activates. Skip just in time when the control is so central that every user hits it in the first thirty seconds. The send button on a prompt in Claude or Cursor needs no tutorial. The core canvas in v0 or Lovable explains itself through immediate output. If your first run still needs heavy explanation after applying smart defaults and agentic flows then fix the product not the tutorial layer. The best teams in 2026 deleted ninety percent of their old onboarding assets and replaced the remainder with these precisely timed hints. Activation jumped. Support volume dropped. Users felt smart instead of instructed.

Just-in-time tutorials turn user curiosity into the only trigger that matters and make every explanation feel like it was the users own idea.