Diligence sample
Production record sample (redacted)
Core claims system + observability - Workflow Deploy. A redacted AP / insurance-ops style production record so buyers can see the structure we ship - criteria, collision map, economics, and sign-off - without mistaking it for a named client case study.
Methodology sample - seeded evidence
This page is a methodology sample with seeded evidence structure. It is not a named client case study, not a performance guarantee, and not a substitute for criteria signed on your real cases.
Related: Evals & evidence
Archetype
Insurance ops - claims exception triage
Package
Workflow Deploy + Registry
Client (redacted)
[REDACTED] - regional payer / TPA (~1,200 FTE)
Decision
Production - acceptance signed
Workflow
Claims exception triage and resolution agent
Duration
Scoped to systems and volume (no fixed calendar promise)
Agent triages tier-1 claims exceptions against the core claims system with human review on amounts and clinical flags above threshold. Owner signed acceptance criteria v1.0; production record v1 committed to client Git; economics attestation filed for finance review.
Engagement path
Opportunity audit
Step 1
Workflow map, systems inventory, manual cost baseline (FTE-equiv / yr)
Build
Step 2
Agent + harness on client eval project; model routing doc; human gates
Production deploy
Step 3
Live on production-candidate claims queue; shadow then limited traffic
Registry
Step 4
Production record v1 + Production Pack + economics baseline
Platform collision
What each system covers - and the gap Falnor closes so production can sign off.
Core claims / adjudication
Covers: Exception queues, benefit rules, approval limits
Gap: Agent approval path required explicit owner rules for high-dollar and clinical flags
Eval / observability harness
Covers: Trace logging, eval runs on real cases
Gap: Does not replace signed acceptance criteria in the production record
Model provider (Azure OpenAI / OpenAI)
Covers: Classification and draft resolution routing
Gap: Documented fallback model and eval-on-upgrade gate in economics block
Acceptance criteria (illustrative)
| Criterion | Test | Threshold | Owner |
|---|---|---|---|
| Resolve or escalate within SLA | Production-sampled exceptions | 95% within agreed SLA - 100% audit log | VP Claims Ops |
| No auto-adjudication above dollar / clinical gate | Harness + manual spot check | Zero violations in shadow | Compliance / Medical Director designee |
| Token spend vs baseline | Monthly economics block | Under agreed % of equivalent manual cost | CFO / Finance Ops |
Economics block (illustrative)
Token spend
Scoped at discovery (illustrative: mid four figures)
FTE-equivalent
Illustrative: hundreds of hours / month on in-scope queue
Net claim
Agent opex under agreed share of manual processing for scoped queue
Typical next steps
Agent Optimize + Operate
Monthly criteria drift + economics refresh + eval-on-upgrade
Workflow #2
Denial management or prior-auth follow-up - reuse production-record playbook