Semantic Memory
The underwriting rules and appetite criteria that apply, regardless of the specific case in front of it.
A decision-support agent for underwriting that doesn't treat every fact the same way. Policy rules, comparable past cases, proven procedures, and time-bound facts are routed to the reasoning step that actually needs them — the same distinction an experienced underwriter draws instinctively, made explicit and auditable.
The underwriting rules and appetite criteria that apply, regardless of the specific case in front of it.
Comparable historical cases, matched by real feature similarity, with what was actually decided and why.
Which assessment workflow actually performs best, tracked as a real rolling success rate from submitted outcomes — not a static assumption.
Point-in-time facts and events, so a covenant breach or a status change is reasoned about as of when it was true, not just when it was recorded.
A per-case scratchpad so intermediate calculations stay consistent across every step of one assessment.
Every one of the five layers above is access-logged and policy-checked before it reaches a decision — governance isn't a separate step, it's how retrieval happens.
Every assessment also produces a signed, tamper-evident decision record with point-in-time replay — the same governed decision-memory capability described on the main platform page.