Summarize an insurance denial letter for an appeal
The poster used an AI assistant to summarize a denial PDF to draft an appeal, but it invented an appeal deadline.
An agent built on Sanity queries real content to tell people in India which government schemes they are entitled to.
An agent built on Sanity queries real content to tell people in India which government schemes they are entitled to.
The agent runs on Inception's Mercury via the Vercel AI SDK and reaches a Sanity Context MCP endpoint in GROQ mode. It extracts the person's details, reads the schema, and makes one GROQ query for schemes with typed eligibility rules. It then checks every exclusion rule. A second pass formats cards as JSON, and the route attaches the apply and source URLs from the dataset.
Works in English and Kannada with follow-ups. The dataset has 125 schemes, but only 10 flagships have structured rules (33 rules, 11 exclusions). A run takes 30 to 45 seconds. The author did not verify every rule line by line.
The first Knowledge Base route burned 20 tool calls without answering. Switching to GROQ mode fixed it, but it needs a deployed Studio.
I am a [age] year old [occupation] living in [state] in India with household income of about [amount] and [other relevant details like caste category, land, or family]. Search official government sites to find central and state schemes I may be eligible for. For each, give the benefit, eligibility, documents needed, and the official application link. Do not submit any applications or enter personal ID numbers.
The poster used an AI assistant to summarize a denial PDF to draft an appeal, but it invented an appeal deadline.
A tenant used ChatGPT and Gemini to review a landlord's apartment deposit dispute, identify legal violations, and avoid losing $4,200.
Poster used Claude over a weekend to plan and build a self-hosted app that organizes receipts for an accountant and replaces paid bookkeeping software.