Training the AI: knowledge base, reply rules, and AI Intelligence
PilotPM's AI improves from three inputs you control.
1. Help Center articles (knowledge base)
Published articles are the AI's source of truth. The more your real workflows and policies are documented, the more drafts are grounded โ which raises confidence and reduces edits. Start with your most common questions.
2. Reply rules
Reply rules capture canonical answers and tone for recurring situations. The AI uses them to stay consistent with how your team actually responds.
3. Learning from edits
Every time a teammate edits an AI draft before sending, PilotPM records the difference. Over time these "deltas" reveal where the AI is reliably right and where it needs more examples.
Tracking progress
The AI Intelligence view (Owner-only) shows your effective automation rate over time, a per-category edit rate, confidence calibration (does "90% confident" actually mean 90%?), and how many learning signals have been captured. Early on this is sparse โ it fills in as real volume flows through the inbox.
Was this article helpful?
Or open the chat in the corner to ask follow-up questions.