The Scenario Harness is a world: simulated claimants, employers, and provider offices that are late, incomplete, and occasionally wrong, driven by scripted scenarios with expected outcomes, under a clock that runs weeks in minutes. It proves every Claimwell product before go-live. It also tests anything else you are about to change.
A scenario names the claimant, the employer, the plan, the condition, the timeline of world events, and what should happen. The simulators play the people, with parameters for how late, how accurate, how complete, and how cooperative they are. The harness runs the scenario and compares the outcome, the money, the dates, the letters, and the human touches against what was expected.
Claimants who reply late or with the wrong date of birth. Providers who fax the same statement twice, or for the wrong patient. Employers who use the reference token, or don't.
Forms rendered to PDF, then degraded per channel: skew, fax headers, missing pages, cover sheets, unrelated mail in the batch.
A world clock with a scale factor. Pause it and step a day at a time in front of a room.
Final state, closure reason, touches by handoff point, benefit amounts against golden outputs, clock dates, required letter content. Zero constraint violations or the run fails.
Your document mix through the Mail Room. Your hardest files through File Intelligence. Every agent against every scenario before it sees a claim.
Widen the straight-through list, lower the duration tolerance, replay last quarter, diff the touches and cycle time before publishing.
Route the extraction task to a different provider and compare outputs side by side on the same documents. Score any vendor's output against the generator's known truth.
A dozen claims in flight, the clock paused. Walk a new examiner through a denial package, or show an auditor exactly what the system does on a bad day.
It is the one product that never touches a production system or a real person. That is why it is the safest place to start a conversation with compliance.
Every release, every strategy version, and every model change is replayed against the full scenario library before it touches a claim. The results are written to your own audit store, next to the evidence behind the warranties. A release that fails any constraint check does not ship.
The reference set covers every state and every handoff point. Your scripted claims sit beside it, and you can add one whenever a real claim teaches you something.
Pass and fail by check, by scenario, by strategy version, by model, for every run. Not a Claimwell portal. The same place an auditor already looks.
A constraint violation anywhere in the run is a hard stop. There is no override for it, on our side or yours.
Extraction and match accuracy by document type and degradation level, run on your document mix before the Mail Room sees a real page.
Illustrative. The same report your platform and compliance teams see for every release, strategy version, and model change.
The harness ships with sandboxed claims administration, HRIS, plan repository, payment, correspondence, and guideline services that behave like the real ones, with the same APIs. It runs from one command on a laptop or in a sandbox account. To test your own operation, point its channels at your test environment instead.
Claims admin with a Guidewire-shaped API, HRIS census feed, plan repository with certificate PDFs, payments, correspondence, guidelines, LTD intake.
Employers, employees, claimants, conditions, plans, and documents, all from seeds. Replays identically.
Eleven reference scenarios covering every state and every handoff point. Script your own in the same format.
Scale, pause, step. Production scale is one.
Route tasks to your own endpoint or a frontier provider for comparison. Simulator calls are tagged so they never count as platform behavior.
Pass and fail by check, by scenario, by strategy version, by model. Exportable.
The Mail Room's accuracy by document type, File Intelligence's delta on a conflicting file, each agent's touches and cycle time. If a product is on this site, the harness has a scenario for it.
We'll script it, run it under accelerated time, and show you where it would have gone to a human.