Claim files that are complete before they leave the hospital.
ClaimReady works at the hospital insurance desk. An AI agent reads the claim, runs the patient's policy rules, links every bill line to a clinical record and fixes what it can. Any change to money waits for the desk. The insurer then re-checks the packet itself and approves clean claims on arrival.
96.7%of discharges decided within 3 h once the file is complete (IRDAI data, Dec 2025)
59%of patients waited 6 to 48 h for discharge approval (LocalCircles, n = 30,366)
8 to 10%of claim payouts lost to fraud, waste and abuse (BCG and Medi Assist, 2025)
Case 8 goal
Settle 60% of standard claims in under 10 minutes. Our simulation models 64% decided in under 10 minutes with ClaimReady, if insurers auto-approve about 75% of clean files. How we modelled it
Cut fraudulent payout leakage by 30%. Our fraud model's base case is a 33% cut (range 20% to 40%); the pilot measures it against control hospitals.
Six demo claims, and where each one needs the agent
Each row is a synthetic claim built around one real failure mode. Pick a row to open it and run the agent.
Claim
Policy rules
Evidence links
Agent fixes
Desk approval
Insurer re-check
Designed outcome
C-103 Imran ShaikhRs 64,200
Hospital code not in package list
Maps code from closed list
Desk approves removing Rs 6,200 kit
Auto-approved after desk approval
C-102 Sunita DesaiRs 42,250
Missing itemised bill
Fetches itemised bill
Auto-approved; patient told about Rs 690
C-106 Kiran PatilRs 28,200
MRI has no record; lab report reused
Flags both; does not fix
Routed to adjudicator
Flagged to adjudicator
C-104 Meena IyerRs 92,900
10% co-pay (C5.4)
Day-3 ICU note missing
Fetches ICU note
Above fast-lane cap
Clean, above Rs 75,000 cap: adjudicator
C-105 Arjun MehtaRs 38,600
Room rent over cap (C3.1)
Tells desk the deduction
Auto-approved; Rs 10,463 deduction disclosed
What you will see in a claim
- The claim file, bill lines and documents
- Each bill line linked to its clinical evidence
- Policy checks with the clause they come from
- The agent's trace: every tool call and its result
- The desk approval gate, when money would change
- The simulated insurer's re-check and decision
Who decides what
Agent
Reads, checks, links evidence, fetches records, maps codes from a closed list
Desk executive
Approves or rejects any change to a billed amount
Insurer rules
Auto-approve clean claims up to Rs 75,000 (rule FL-1)
Committee
The only party that can deny. The agent never denies.
What is real in this demo
Real
The agent loop, 12 tools, the policy rules engine, the approval gate, the audit trail
Simulated
Hospital records system, insurer decision service and its response time
Synthetic
All six patients, documents, the hospital, insurer and policy
Modelled, not shown
The 64% decided in under 10 minutes is a simulation result, not measured here