A mid-sized regional property & casualty insurer, ~1.2M active policies
Financial Services / Insurance
The shape of this problem
The same judgment, made differently every time
The problem
Claims arrived as unstructured attachments - PDFs, phone-photo scans, handwritten forms - into a shared inbox, where 14 clerks read, classified, and keyed each one by hand. First-touch to adjuster assignment averaged 4.3 business days, and roughly 9% of claims were misrouted at least once.
What we built
We began with a two-week audit rather than a model, sampling 1,800 submissions to map where time actually went. Two findings reshaped the scope: 60% of volume came from four claim types, and most misrouting traced to a data-entry step, not a reading problem. We redesigned the routing workflow first, then built NLP extraction against the four high-volume types, running it in shadow mode until accuracy held above 94% on a live holdout.
What moved
| Measure | Before | After |
|---|---|---|
| First-touch to adjuster assignment | 4.3 days | 7 hours on 72% of volume |
| Classification accuracy | no prior measure | 96.8%, versus a 91% human baseline |
| Misrouting | 9% | 1.4% |
Misrouting
Figures are from the practice's own delivery records for the engagement named, measured against the process that preceded it.
We expected an AI tool. What we got first was an honest map of why our intake was slow - and fixing that before the model went in is why the numbers held.Head of Claims Operations
- Service line
- Document & Request Triage
Start here
Which of the four is yours?
Tell us the documents and the monthly volume and we will send the two closest records, with the figures behind each and what it would take to repeat them on your material.
You get a reply within one working day, from the engineer who would do the work.