Multi-User AI Email Triage Assistant for Gmail
Gmail-integrated AI triage system that prioritises email, extracts deadlines and drafts FAQ-grounded replies, using a rules-first engine with an LLM fallback.
Business Challenge
High-volume inboxes bury urgent messages and deadlines, and manual triage wastes time. Naive AI triage tools call the Gmail API and an LLM for every message, which is slow, costly and hits API quotas, and they rarely isolate data between multiple users.
Solution Delivered
Built a multi-user triage platform with a FastAPI dashboard, a Google Apps Script Gmail add-on and a CLI. Messages are classified rules-first, with a Groq LLM fallback for ambiguous cases. The system learns adaptively from user corrections, scores priority, extracts deadlines and provides local semantic search with FastEmbed. Authentication uses OAuth 2.0 (PKCE) with per-user data isolation. FAQ-grounded replies are saved only as Gmail drafts so a human always reviews before sending. Gmail API usage is batched.
Outcomes Achieved
Reduced Gmail API calls by 92.5% (from 200 to 15 per inbox) across 1,800+ messages through batching. Delivered automated prioritisation and deadline extraction with human-in-the-loop draft replies, backed by 59 regression tests and documented validation against real Gmail inboxes.
Measurable Business Outcome
Gmail API calls reduced by 92.5% (200 to 15 per inbox) across 1,800+ messages; 59 regression tests passing.