The challenge
Claims volume had grown 31% over three years; headcount had grown 4%. Every adjuster was working out of a single queue that mixed simple motor windscreen claims with complex bodily-injury cases. The simple cases got slow service; the complex cases got rushed service. Neither outcome was acceptable to the leadership team, and neither could be fixed by adding more people fast enough.
Discovery
A two-month claims audit segmented the queue into three real populations: 47% straight-through candidates, 41% routine adjuster review, 12% complex multi-party cases. The audit also showed that the triage decision currently being made by the most senior adjusters could be made within seconds by a model trained on the insurer's own historical claim records — freeing those adjusters for the complex 12% where their judgement actually mattered.
Our approach
Trained a triage model on three years of historical claims with the outcome ledger attached, producing per-claim category, complexity score, and recommended routing
Routed straight-through cases into an automated workflow with explicit stop conditions — any anomalous signal kicks the case back to a human adjuster within seconds
Surfaced model reasoning to the adjuster reviewing each routed case, so the team can challenge and improve the model rather than blindly accept it
Established a weekly model-review meeting between the chief claims officer, the data team, and the compliance lead — keeping governance ahead of throughput
The Outcome
A triage layer at claims intake that segments the queue into three flows: straight-through, routine adjuster, and complex. The senior adjusters spend their time where it matters; the routine cases get faster service; and the model improves week over week from the adjusters' own challenges.
The results
“Our senior adjusters stopped triaging windscreen claims and started spending their time where their judgement actually mattered. That is what changed claimant experience.”
