The Adjuster Was an Algorithm: The Policyholder’s Bad-Faith Case Against AI Claim Handling

Insurance bad faith has always assumed a human adjuster who investigated the claim, evaluated it, and decided it. Insurers now use AI for automated triage, severity scoring, settlement recommendations, fraud detection flags, and claim denials and reductions. This session starts from one premise: an insurer cannot outsource its duty of good faith to a machine, and when an algorithm does the adjuster’s job, the algorithm becomes evidence of how the insurer handled the claim. It uses the Oklahoma AG allegations as a roadmap — that insurers used AI systems to undervalue claims, make settlement offers below fair value, increase claim closure rates, and put profit ahead of accuracy — and explains why a government investigation helps a private case: it creates a factual narrative, makes internal documents discoverable, and identifies industry practices. It then maps that conduct onto the traditional bad-faith elements — unreasonable claim handling, lack of proper investigation, and knowing or reckless disregard of the insured’s interests — through four theories: biased inputs producing a biased investigation, algorithmic undervaluation, institutionalized claims minimization, and failure to supervise the AI. It closes by tracing the decision chain from data ingestion through scoring to the claim outcome, so you can show why a claim was denied, delayed, or underpaid, and by putting the reasonableness of the insurer’s reliance to a jury.

Ask for the algorithm and the insurer will call it proprietary technology, a trade secret, or the vendor’s — and say none of it matters because a human made the final decision. This session treats the algorithm as fact evidence rather than technology: it took part in handling the claim, influenced the valuation, and generated recommendations, so it belongs in the claim file. It works from one principle — show me everything that influenced the claim decision — and frames requests in four categories: system architecture, claim-specific AI data, performance validation materials, and corporate knowledge. It sets deposition topics for the four witnesses who matter: claims staff, the data science team, compliance, and the vendor. It answers trade-secret objections by showing the insurer put the algorithm at issue, and by offering attorneys’-eyes-only designation, source-code review protocols, independent expert review, and confidentiality orders, with production limited to the claim. It ends by assembling the AI claims file itself: adjuster notes alongside AI-generated recommendations, risk scores, valuation calculations, automated alerts, override logs, internal appeals, and model-generated communications.