Auto AI-Drafts lets Productlane's AI automatically generate a draft reply whenever a new customer thread arrives or a new inbound message is received. Instead of clicking "Draft a reply" each time, the AI proactively prepares a response in the compose editor so your team can review, edit, and send it immediately.
The AI uses the full thread context along with your docs, changelog, roadmap, resolved threads, GitHub PRs, and connected Postgres databases to ground its response.
AI Chat must be enabled — Auto AI-Drafts builds on the AI Agent, so the agent must be turned on in Settings → AI Agent first.
Pro plan or above — this feature is available on Pro and Scale plans.
Go to Settings → AI Agent.
Scroll to the Auto AI-Drafts section.
Turn on the Enable Auto-Drafts toggle.
Choose when drafts are generated:
Every inbound message (default) — a draft is generated for every new inbound customer reply, as well as new threads.
New threads only — drafts are generated only when a brand-new thread is created. Follow-up messages from the customer do not trigger a new draft.
Optionally provide free-text instructions that guide the AI's drafting style and content. Use this to encode institutional knowledge, preferred phrasing, or topic-specific guidance (e.g. "For billing questions, always link to Settings → Billing rather than sharing Stripe URLs").
There is also a Suggest custom prompt button that analyzes your recent resolved threads and auto-generates a custom prompt based on patterns it finds in your team's past replies. This requires at least 11 answered threads.
When a draft is being generated, a "Drafting…" badge appears in the reply composer. Once ready, the draft loads into the editor and the badge updates:
Drafting… — the AI is still generating the draft (animated indicator).
Use AI draft — the draft is ready and available to apply. Click the badge to load it into the composer. This state appears when the composer already has content (e.g. you started typing), so the draft won't overwrite your work without your consent.
Drafted by AI — the draft has been applied to the composer. Hover the badge to see which sources the AI consulted.
Draft failed — the draft could not be generated. Hover for details. You can still write your reply manually.
Hovering the badge reveals the sources the AI consulted while drafting. Each source is clickable and links to the original item:
Docs — help-center articles
Changelog — changelog entries
Roadmap — roadmap projects
Threads — previously resolved threads
GitHub PRs — pull requests from your connected repository
Postgres queries — results from your whitelisted database queries
Empty composer — the draft is automatically loaded into the editor. No action needed.
Composer has content — the draft appears as an opt-in "Use AI draft" button so it never silently overwrites what you've typed.
Editing an applied draft — as soon as you type into a draft that was applied by the AI, the "Drafted by AI" badge disappears and the draft is treated as your own.
Sign-off — the draft uses a placeholder for the agent name that is resolved to whichever team member reviews and sends the reply.
Drafts are automatically cancelled in these situations to prevent stale content:
A new inbound message arrives on the same thread while a draft is being generated.
A new thread-created event fires for the same thread.
You send a reply on the thread — the draft is cleared after sending.
Keep your docs up to date — the AI drafts are only as good as the sources it can consult. High-quality, current documentation leads to better drafts.
Use the custom prompt — encoding your team's institutional knowledge (common answers, preferred links, things to avoid promising) significantly improves draft quality.
Try "Suggest custom prompt" — if you're not sure what to write, let the AI analyze your past replies and suggest instructions for you.
