Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.buildbetter.ai/llms.txt

Use this file to discover all available pages before exploring further.

Projects ships several AI surfaces. Each one is grounded in your real workspace data — signals, customers, PR activity, Linear tickets, and call transcripts — and most show you the evidence behind every claim.

AI Roadmap (List Page)

The banner at the top of /projects clusters your active and recently shipped projects into themes with a one-sentence narrative. See List & Views for the full controls. What it does:
  • Pulls projects from the past 30, 90, or 365 days (your choice)
  • Asks an LLM to cluster them into themes (e.g., “Search relevance”, “Onboarding friction”)
  • Returns a one-sentence narrative summarizing the roadmap
Two presentations:
  • Themes view — three columns (Scoping · next up / In flight / Shipped) with up to 6 projects each
  • Timeline view — gantt-style swim lanes grouped by AI themes, product areas, or initiatives
Click Refresh to regenerate the clustering from the latest project state.

Project Rollup (Detail Rail)

An AI synthesis of the project, written in sections: What / Why / Who / How / Risks / Next actions. Rendered as a two-column grid with one-click Copy and Refresh. Evidence: the rollup reads only approved signals attached to the project. Anything in the “suggested” review state is ignored until you approve it. Header: “Approved evidence only · [generated timestamp]” — so you can tell at a glance whether the rollup is current.

Suggested Signals Banner

Auto-fires on the detail page when a project has fewer than 5 attached signals and the title is at least 6 characters.
  • Searches your workspace’s signals using the project title as the query
  • Defaults to external / customer boundary
  • Returns up to 5 matches with match %, summary, person/company, and a boundary tag
  • Multi-select with checkboxes; Add N to review attaches them as “suggested” status (not auto-approved)
Use it as a quick way to seed evidence early in a project’s life — and then approve the signals that actually fit.

Similar Projects Banner

Detects potential duplicates using cosine-distance similarity on project embeddings.
  • Surfaces up to 3 candidates with a match % ((1 − distance) × 100)
  • Merge collapses the candidate into the current project — moves its signals over and archives the duplicate
  • × dismisses the banner for the session
Run this on any project you suspect overlaps with an older one — it catches duplicates that wouldn’t share keywords.

AI Status Suggestions

Lives inside the Status at a Glance card. When a real project has zero blockers and zero next actions, a Suggest status button appears. The agent returns three buckets:
BucketWhat It Surfaces
Potential blockersEnvironmental issues the agent infers (e.g., “waiting on infra team”, “design review not started”). Includes reasoning.
Next actionsSpecific actions to advance the project. Includes reasoning.
Recently doneWhat looks already done from PR/signal evidence, in case it’s worth marking complete.
Suggestions render as ghost/muted rows. Click any one to promote it to a real blocker or action. If the card already has blockers or actions, the AI panel stays out of the way.

AI Routing Suggestions (Triage → Projects)

When a tracked item is in triage, the AI route banner can suggest attaching it to a high-confidence existing project or creating a new one. Accepting the suggestion promotes the item to /projects. See AI Routing in Triage for the full flow.

Post-Release Monitor

Only appears for projects in the Released stage. The monitoring agent watches signals received after release and flags regressions.
  • Tracks the post-release signal window (default 90 days)
  • Clusters signals by sentiment (positive / neutral / negative)
  • Alerts at three severities — info, warning, critical — when spikes in negative or support-related signals look like a regression
  • Auto-advances the project to Stable after 90 days with no regressions

What the AI Doesn’t Do Without Your Approval

  • It never auto-attaches a signal to a project. Suggestions land in the “suggested” review state and require your approval.
  • It never auto-merges duplicates. You always click Merge.
  • It never silently changes status. Stage-advance suggestions appear as a “suggested” pulse on the stepper, but you make the call.
  • It never sends a close-the-loop update. The composer is always reviewed before sending.
If you want to see what evidence drove a particular AI suggestion, expand the rollup or open the related signals — every claim cites the underlying customer signals.