Works from the case record
The agent reads the inputs, data and decisions recorded for the case, not general knowledge about insurance.
Investigate real cases with the context behind them.
An investigation agent helps authorised users inspect case records, relevant rules and recorded decisions. It can suggest why an outcome occurred and what to review next, giving domain experts a practical starting point for improvement.
How an investigation works
From a real question about a case to a change someone has reviewed. The agent prepares; people decide.
Explanations you can check
An agent's explanation is a starting point. It must be checked against the source data and the rules that ran.
Shows its evidence
Each explanation points to the records and rule results it is based on, so the owner can follow it.
Proposes, never decides
Rule changes the agent suggests go to review. People decide whether a change applies.
Limited to authorised users
Only people with the right role can investigate a case or propose a change.
What changes for rule owners
A practical starting point for improving the process, grounded in real cases rather than assumptions.
A clear starting point
Owners start from a proposed explanation and its evidence, instead of piecing the case together by hand.
The same context
The agent sees the same case record as the advisor and the underwriter who handled it.
Changes with reasons
A proposed rule change carries the case and the reasoning behind it into review.
Improvements from real cases
Rule changes start from outcomes your teams actually saw, not from guesswork.
Questions, answered
How the investigation agent works, and how it is kept under control.
It helps authorised owners examine why a case came out as it did. It gathers the case inputs, the data fetched, the rules that applied and the decisions recorded, then suggests an explanation and what to review next.
No. It can propose a rule change, which goes to review. People decide whether a change applies, and decisions on cases stay with the advisor or underwriter.
Treat it as a starting point. AI can make mistakes, so the explanation shows the evidence it used and should be checked against the source data and the rule results before anyone acts on it.
Only users with the right role, such as the people responsible for the underwriting rules. Advisors and underwriters do not need it to handle cases.
Start with one case.
We can walk through an investigation on the property risk assessment solution, then discuss your own process.