Documents Belong in the Core

The paperwork is never going away. What changes is who has to touch it and how many systems it moves through before it turns into a decision.

When we designed the Mercury Policy and Claims Administration System, we kept running into the same pattern at carriers we talked to. The policy admin or claims system was the system of record in name only. The real work of reading documents, pulling out key fields, and deciding what mattered was happening in a separate document management tool, a separate OCR service, a separate fraud-scoring product, and a separate folder on somebody's shared drive. Every handoff lost context, and every handoff added risk.

We took the opposite approach. Document imaging lives inside Mercury. Natural language processing of unstructured attachments lives inside Mercury. AI-powered document fraud detection, which assigns a one-to-one-hundred rating with explainable per-document features, lives inside Mercury. When a new attachment lands on a policy or claim, it is read, classified, indexed, and scored in the same file the underwriter or adjuster is already working in.

That sounds like a small integration convenience. In practice, it changes the operating model in four ways.

First, it compresses cycle time. When an adjuster does not have to open three tools to triage a medical record, the work gets done in minutes instead of an hour. Multiply that across a claim file with fifty attachments and the math becomes obvious.

Second, it closes the audit loop. Every document extraction, every classification, every fraud score is recorded against the file with a timestamp, a model version, and the features that drove it. When a regulator or a reinsurer asks why a decision was made, the answer does not live in someone's email archive.

Third, it reduces the surface area that has to be kept compliant. One platform with role-based access controls and a single audit log is easier to certify and easier to defend than five integrated but separately governed tools. HIPAA, state privacy statutes, and market-conduct examinations all benefit from a single place to look.

Fourth, and maybe most importantly, it changes what the people doing the work can spend their attention on. Underwriters and adjusters are not paid to be document clerks. They are paid for judgment. When the platform does the reading, the classification, and the fraud signal generation, the human gets to do the judgment part earlier in the file and with better information.

I am deliberately not claiming that Mercury writes narrative summaries with large language models, or that it drafts correspondence on behalf of adjusters. Those are genuine industry opportunities, but they are not what we ship today. What we do ship is the foundation: imaging, extraction, classification, indexing, and fraud scoring, all native to the core platform, all auditable, all delivered by Quick Silver Systems with the implementation discipline our clients expect.

If you are evaluating a core system right now, ask where document intelligence lives. If the honest answer is that it lives in a product your vendor did not build and cannot fully explain, that is a signal worth listening to.

-- Sean Pitcher, CEO, Quick Silver Systems, Inc.