Project 10 · AI operations assistant
An operations assistant that reads shared inboxes, classifies each request and routes it to the right team, with a person reviewing before anything important happens.
- Market
- India
- Year
- 2026
- Timeline
- 10 weeks
- Team
- 4 people · PM, 2 engineers, ML engineer
- Delivered
- AI agents · Integrations · API
01
The challenge
Operations teams lose hours sorting shared inboxes by hand. Requests sit unread, go to the wrong team or get answered twice. Fully automatic tools are risky when a wrong action costs money or trust.
02
Our approach
We built LLM agents in Python that classify requests and draft next steps, using retrieval over the team's own documents. Every action goes through a review step that people can approve, edit or reject.
- 01
Inbox reading
Connects to shared inboxes and pulls out the request, the sender and any key details.
- 02
Request classification
Each message is labelled by type and urgency using the team's own categories.
- 03
Routing to teams
Work is sent to the right team's queue or tool, with a short summary attached.
- 04
Human review
Suggested actions wait for approval, and every decision is logged so the system can be checked.
Platforms
- Integrations
- Review console
- API
Technology
- LLM agents
- RAG
- Workflows
- Python