Eluna

(01) Problem

Businesses needed to automate multi-step processes involving documents, emails, and tickets. Sensitive steps required human approval, while routine actions needed autonomous execution in business systems.

(02) Approach

1. 01Implemented AI agents to extract data and plan workflows from incoming inputs.
2. 02Integrated RAG over company knowledge to retrieve relevant policies for decision-making.
3. 03Designed a human-in-the-loop approval mechanism for steps requiring manual verification.
4. 04Executed actions such as ERP bookings and notifications based on agent plans and approvals.

(03) Architecture

(04) The hard part

The hardest part was determining when an agent could act autonomously versus when it must request human approval. This was handled by implementing strict rules and maintaining a complete audit log for every step.

(05) Result

The platform runs in production, automating end-to-end business processes.