Review insurance claims with DeepSeek, forms, and Telegram

Quick overview

This workflow collects medical insurance claims via an n8n form, validates them against policy data in n8n Data Tables, and either auto-approves/rejects or sends an AI-drafted recommendation to a senior reviewer on Telegram for approval, rejection, or revision, then logs and notifies the final decision.

How it works

1. Receives a medical claim submission from an n8n Form Trigger.
2. Looks up the submitted policy number in an n8n Data Table and applies rule-based eligibility checks to calculate deductible, remaining coverage, and maximum allowed payout.
3. Automatically approves low-risk claims under the configured limit or automatically rejects invalid/ineligible claims based on the rule results.
4. For claims requiring human review, uses DeepSeek (via the LangChain agent) to draft a recommended payout, decision note, and claimant message, then enforces a guardrail cap so the payout never exceeds the rule maximum.
5. Sends the claim summary and AI draft to a senior reviewer in Telegram with a form to Approve, Reject, or Request changes, and loops up to three revision rounds when changes are requested.
6. Builds a final decision record, logs it to an n8n Data Table, and sends the outcome and claimant message to Telegram.

Setup

1. Create and populate two n8n Data Tables (one for policies and one for decisions) and update the Data Table IDs in the policy lookup and decision logging steps.
2. Add a Telegram credential, then set the target Telegram chat ID used for the review request and final notification messages.
3. Add a DeepSeek API credential for the LangChain chat model used to draft the senior-review decision text.
4. Review and adjust the rule parameters in the code (for example the auto-approve limit, covered types format, and deductible/coverage fields) to match your policy schema and business rules.