ICUA is the seven-stage methodology Interpose uses to take a submission from a raw Countervail scan through to a bound underwriting decision. Each stage turns one question into a structured answer, and each answer feeds the next stage.
The stages show up as a strip across the top of the underwriting workbench. This page explains what each one means and why it sits where it does in the sequence.
We start by looking at the AI systems in use: which vendors are involved, what records they hold, what gets preserved, and who would need to produce what if a subpoena arrived.
Before anyone can price the risk of compelled production, we need a shared picture of what records exist and where they live.
Based on the substrate, each account gets placed on a five-level scale from IC-1 (lowest exposure) to IC-5 (highest). The class reflects how much record-custody risk the account carries, not whether the AI itself is working correctly.
Exposure classes make accounts comparable across the book and let conditions and premium scale with actual custody surface, not guesswork.
We set the coverage triggers — the specific events that would cause a claim to attach. Every account gets a primary trigger (compelled production of retained records) plus any secondary triggers the substrate warrants.
Triggers make explicit what is and is not covered, so there are no surprises at claim time.
We set the conditions, warranties, sublimits, and exclusions that go on the policy. Most come from the rules engine based on the substrate; some are added manually by the underwriter.
Conditions match the policy shape to the actual risk pattern rather than defaulting to one-size-fits-all terms.
Critical underwriting facts either get confirmed, or the underwriter sends an RFI (request for information) back to the broker. The submission cannot bind until the critical facts are confirmed.
Some facts move the decision materially; those need to be verified before a quote goes out.
The workbench produces one of four decisions: accept, condition, refer, or decline. Drift between the scan at submission and the scan at renewal can reopen a completed decision.
The decision is the output the broker sees. Drift detection keeps it honest across renewal cycles.
A human underwriter reviews the machine-suggested decision and either accepts it as-is, modifies it, or finalizes it with notes. Nothing binds on machine output alone.
The machine does the arithmetic. The human does the underwriting.
The stage strip is visible on every submission inside the workbench and progresses as the submission moves. A submission that reaches stage 7 (sign-off, human finalized) is ready to bind. A submission that stalls on stage 5 (fact verification) is waiting on an RFI response from the broker.
ICUA is documented in full in the underwriting memo export. When a submission produces a memo, the methodology footer lists which stages are complete, which are pending, and where the current submission sits in the sequence.