Onboarding-free product UX (in-product AI agent answers everything)
Wizard onboarding has 30-60% drop-off. An in-product agent that knows the user's actual state cuts time-to-value from days to minutes.
What
Replaces traditional onboarding flows (wizards, tooltips, doc-tours) with an in-product AI agent that answers any question contextually — "how do I add a refund?", "where's my payout schedule?", "what does this column mean?". The agent has access to the user's actual workspace state (their accounts, their data, their permissions) and answers with citations to docs + actionable buttons.
Why it matters
Traditional onboarding has 30–60% drop-off. An in-product AI agent that knows the user's state cuts time-to-value from days to minutes. A payments group cut merchant onboarding from 180 minutes → 30 minutes by replacing the wizard with an agent. At several-thousand-client scale this was ~$300K/year + 10–15 onboarding FTE freed.
End-to-end
- Inventory current onboarding artifacts (wizard, tooltips, help-center articles, demo videos).
- Consolidate into the unified KB (#76) — agent will query this.
- Embed the agent UI in-product — a persistent chat panel that's aware of which page the user is on.
- Wire context: current user, current workspace, current permissions, page being viewed.
- Give the agent action capabilities — not just answers but buttons like "I'll set this up for you" with explicit confirm.
- Track time-to-first-value (TTFV) before/after — should drop 4–10×.
- Phase out the wizard: agent handles new-user questions; wizard becomes a fallback for unauthenticated tours.
Gotchas
- The agent must have action capabilities, not just Q&A. An onboarding-free flow that only answers questions still leaves users to navigate menus. The compounding value comes from "I'll do this for you" → user confirms → done.
Tools
- Unified KB (#76)
- An in-product chat UI surface
- LLM with action/tool-calling (Claude with tool use, GPT with function calling)
- API for the agent to read user state + take actions