Claims lifecycle · Stage 04 of 08 · Evidence & claim file
AI claim file review: a decision-ready file, not a summary
Every record read, the timeline rebuilt, the contradictions surfaced, each line citing its page.
In short
Hesper AI is an AI claims resolution platform for P&C carriers, TPAs, and MGAs. Its agents read every document on a claim, medical records, statements, invoices, reports and photos, then produce a decision-ready file: what happened, what the evidence supports, what contradicts it, and what is still missing, with every line citing the page it came from. The adjuster decides.
What is the difference between a summary and a decision-ready file?
A summary shortens documents. A decision-ready file answers the questions the adjuster actually has to settle: is the treatment consistent with the reported mechanism of injury, do the statements agree with the physical evidence, does the invoice match the estimate, what is missing before this can be paid or denied.
The test is simple. If the reviewer still has to open the underlying records to trust the output, the tool moved work rather than removing it. That is why every assertion Hesper writes points back to its page, which is also what California asks of the claim file itself: enough detail that the events and their dates can be reconstructed (10 CCR 2695.3).
What does the file-review agent do?
- Classifies and de-duplicates everything in the file, including scans, faxes, photographs and re-sent copies.
- Reads medical records, bills, statements, reports, estimates and correspondence, rather than only the structured fields.
- Rebuilds the timeline: date of loss, first treatment, gaps in care, return-to-work dates, repairs and payments, in order.
- Surfaces inconsistencies between documents and between documents and the reported facts.
- Runs document forensics on anything that looks altered, synthetic or machine-generated.
- Writes the file with citations, and lists what is still outstanding.
Where does the file go next?
Into the decision. The same evidence supports the coverage position, the investigation if something is wrong, the settlement valuation, and the subrogation demand at the end of the claim. It is assembled once and used at every stage.
That is the difference from a document tool that sits beside the claims system: the file is not an artefact, it is what the next stage decides on.
Do we have to move our documents somewhere new?
No. Hesper connects to your claims system of record through its API and works on the documents already attached to the claim. Data is encrypted in transit and at rest, retained for the subscription term unless you instruct otherwise, and never used to train models. We are SOC 2 Type I attested.
Records-heavy lines feel this first: workers' compensation and liability.
What does Hesper do, and what does your team decide?
| Hesper's agents | Your team |
|---|---|
| Reads and classifies every document on the claim | Decides what to request and from whom |
| Rebuilds the timeline and surfaces contradictions | Weighs the contradictions |
| Cites every assertion back to its page | Spot-checks the citations |
| Lists what is still missing | Decides to pay, deny, investigate or negotiate |
What does the output look like?
- File
- Medical records from four providers, employer report, recorded statement, wage records
- Timeline
- Injury reported on a Monday for a Friday incident; first treatment eleven days later
- Consistency
- Mechanism described in the recorded statement differs from the one in the initial intake note
- Billing
- Two providers billed the same therapy dates; one is a duplicate of the other
- Missing
- No return-to-work note after the second independent evaluation
Evidence cited
- Intake note and recorded statement quoted side by side, both cited
- Duplicate therapy dates shown with both bill pages
- Treatment gap plotted on the reconstructed timeline
Recommended next action: Adjuster requests the missing work note and decides whether the statement conflict needs an investigation.
Frequently asked questions
What is AI claim file summarization?
Software that reads the documents on a claim and produces a structured account of what they say: the timeline, the findings, the contradictions and the gaps. Good implementations cite the source page for every line so the reviewer can verify without reopening the whole file.
How is this different from a medical record summary tool?
Record summarizers stop at the records. Hesper reads the whole file, including statements, invoices, reports and photos, checks them against each other, and hands the result to the next stage of the claim rather than producing a standalone document.
Can the adjuster trust the output without reading the records?
Every line cites the page it came from, so verification is a click rather than a re-read. Adjusters keep the decision, and the agent flags what it could not read confidently instead of guessing.
What about altered or AI-generated documents?
Document forensics runs as part of the review: tampering, synthetic documents and machine-generated text are flagged with the reason, and anything suspicious routes into an investigation-grade workup.
Where is our claim data stored?
Encrypted in transit and at rest, retained for the duration of your subscription term unless you instruct otherwise in writing, and never used to train models. Hesper is SOC 2 Type I attested and GDPR-ready.
Other stages of the claim
Book a demo of Hesper's file-review agents
Bring a real claim file. We will run it and show you the citations.
Book a demo