Coverage Check Agent · Use case
How can an insurer use an AI agent to check a claim against the policy wording on its own infrastructure?
A coverage check agent sets a loss against the policy as written. It finds the clause that grants cover, the exclusions, the limits, the deductible and the endorsements that apply, and quotes each with its clause number. Where the wording is silent it says so. The adjuster decides whether the loss is covered.
One loss.
Sixty pages of wording.
- ✗ 01To check cover an adjuster reads the schedule, the wording and every endorsement, and holds them against the facts of the loss. On a complex policy that is most of the first day.
- ✗ 02A wording cannot be pasted into a public AI service, and a tool that guesses at cover has made the insurer a promise.
Where it stops for a person
Always. Cover is the adjuster's decision.It quotes the clauses.
The adjuster decides.
Step · 01
An adjuster gives the agent a claim.
Step · 02
The agent reads the claim record and finds the policy, its schedule and its endorsements.
Step · 03
It sets the loss against the wording: the insuring clause, each exclusion that could apply, the limit and the deductible.
Step · 04
It writes what it found, each point with the clause quoted and numbered, and lists what the wording does not answer.
Step · 05
The adjuster reads it and decides cover.
Reads from · connectors
Your own APIs
Your wordings and endorsements, indexed once as a knowledge source with each reader's permissions kept. The claim record, through your claims system's own API: you give Katonic the API's OpenAPI description and each operation becomes a tool.
Acts through · MCP servers
Nothing. It changes no claim and sends nothing.
What you set up
- Index the wordings by product and version.
- Import the claims API's description and switch on the read operations only.
- Test against closed claims whose outcome you know.
Where it runs
On your own infrastructure, with a model you serve. Wordings and claims stay there.
How you get it today
This page is labelled Use case: the agent is built in Katonic Studio. You describe the job, connect what is named above, add test questions and release the agent once it passes them. When the Catalog holds a tested template for this job, this page will carry the Template label and show how to take it.
Does the agent decide whether a loss is covered?
No. It sets out what the wording says, clause by clause. The adjuster decides.
What does it do when the wording does not answer the question?
It says the wording is silent on that point and stops there. It does not infer cover.
Does it read endorsements as well as the base wording?
Yes, where they are indexed with the policy. It names the endorsement each point came from.