Field Documentation Evidence Standard

Documentation requirements are not uniform across claim types. A water loss demands documentation of origin, cause, water category, water class, duration, and mitigation status. A wind/hail loss demands slope-by-slope roof evaluation, consistency checks, and soft metal documentation. A tool that applies a generic framework across all claim types will produce generic output — and miss claim-type specific gaps. This capability requires that a tool recognize claim type and activate the appropriate depth module.

A claim is a process, not a moment. The documentation that matters at initial contact differs from what matters at coverage determination, and differs again at close. A tool that captures only the inspection snapshot is documenting one point in the claim, not the claim itself. This capability requires that documentation account for lifecycle position: where the file sits in its arc, what has been established, and what remains open.

Recording what is present is not the same as identifying what is missing. Gap validation means actively surfacing the documentation gaps — the information that should be in the file and isn't. A transcription tool records what the adjuster said. A documentation tool tells the adjuster what they didn't say that they needed to. This capability most sharply distinguishes a documentation tool from a writing assistant.

AI tools produce confident output. Confident output is not the same as accurate output. Left unconstrained, a documentation tool can invent details it was not given, state coverage positions the adjuster has not reached, or present reported information as confirmed observation. This capability defines what the tool must not do: no invented facts, no coverage determinations, no elevation of hypothesis to conclusion, no blurring of observed versus reported versus vendor-documented. A tool that violates these constraints creates liability, not documentation.

A claim narrative is only as valuable as its traceability. Where did each statement come from — the adjuster's original notes, a clarifying question answer, or existing file documentation? This capability requires that the documentation chain be preservable: raw notes to clarifying questions to answers to final narrative. Without provenance, the narrative is a finished product with no visible production record. With it, every assertion can be traced back to its source.

Individual documentation events produce individual files. Aggregated, they produce intelligence. Are the same categories of information missing claim after claim? Are coverage evaluation gaps clustered in a particular claim type? This capability converts documentation activity into documentation intelligence — and makes the FDES useful at the organizational level, not just the individual file level.