PM-JAY AI Challenge 2026: how we won with AI claim and bill verification
The NHA PM-JAY AI Challenge 2026 was a hackathon run by the National Health Authority. BiltIQ entered Problem Statement 2, claim and bill verification, with Agent OS agents that check a claim against the rules and cite the page behind each finding. BiltIQ won the challenge at the IISc Bangalore finale on 8 May 2026.
We'll walk you through it on sample data.
The challenge: verify health claims and bills
Health-claim payouts in India are lost to fraud, waste and abuse, and claims are still checked largely by hand. The National Health Authority ran the PM-JAY AI Challenge 2026 as a hackathon, and Problem Statement 2 asked teams to verify claims and bills.
What we built: AI claim and bill verification with Agent OS
- Read. Agents read the claim documents and the bills.
- Check. They test each claim against the rules set for the challenge.
- Cite. Each finding cites the page and the text it rests on, so a reviewer can check it against the source.
- Decide. In Agent OS, findings go to a person, who accepts or rejects each one. Irreversible and high-consequence actions pause for human approval.
It ran on Agent OS, the same SI agent platform we install on a buyer's own servers.
The result: PM-JAY AI Challenge 2026 won, IISc Bangalore finale, 8 May 2026
BiltIQ won the NHA PM-JAY AI Challenge 2026 at the finale at IISc Bangalore on 8 May 2026. The declaration from NHA is on file.
What this case study does not claim
- It was a hackathon, not a contract.
- We publish no accuracy, speed or volume figure from the challenge.
- We report no deployment data.
What it means for claims teams
The same SI agents can check claims against your policy and bills on your own servers, with outbound calls blocked by default. We prove it on your past claims in a 4–6 week paid pilot, ₹5–8 L, credited against the licence, measured against your manual results. NDA first; on your server or a BiltIQ loaner, your documents stay on site.
Key facts: PM-JAY AI Challenge 2026 case study
- Winner, NHA PM-JAY AI Challenge 2026 (hackathon), Problem Statement 2, claim and bill verification.NHA declaration, May 2026
- Finale at IISc Bangalore, 8 May 2026.NHA declaration, May 2026
- Built on Agent OS: agents read, check and cite the page; findings go to a person.BiltIQ code review, Oct 2026
- A hackathon, not a contract; no accuracy figure and no deployment data are published.BiltIQ
FAQ
What was the NHA PM-JAY AI Challenge 2026?
It was a hackathon run by the National Health Authority around PM-JAY, India's public health-insurance scheme. Teams took on set problem statements; Problem Statement 2 was claim and bill verification. The finale was held at IISc Bangalore on 8 May 2026, and BiltIQ won it. It was a competition, not a procurement or a contract.
What did BiltIQ build for AI claim verification?
We built the entry on Agent OS, our SI agent platform. Agents read the claim documents and bills, test each claim against the challenge's rules, and cite the page and text behind every finding. Findings go to a person, who accepts or rejects each one; consequential actions pause for approval.
Is BiltIQ's AI claim verification deployed at NHA?
This page reports the PM-JAY hackathon win, not a deployment or a contract. We publish no deployment data and no figures from the challenge. What we can offer a claims team is the same Agent OS on its own servers, proven first in a 4–6 week paid pilot on its own past claims, measured against its manual results.
Can insurers or TPAs use the same AI for claims checks?
Yes, as an Agent OS deployment on your own servers. Your SI agents check claims against your policy wording and the bills, and cite the page behind each finding for your reviewers. Start with a paid pilot of 4–6 weeks on your past claims, ₹5–8 L, credited against the licence. NDA first; documents stay on site.
How accurate was the AI in the PM-JAY challenge?
We do not publish an accuracy figure from the challenge. A hackathon result on its own test set does not tell you how the system will perform on your claims. That is why every BiltIQ engagement starts with a paid pilot on your own past files, reported as caught, missed and time per file against your manual results.
Test the same SI agents on your own past claims
We'll walk you through it on sample data. Or write to [email protected], or call or WhatsApp +91 89868 60088 (messages go straight to our CRM).