Evaluating AI-Assisted Claims Review Across Five Lines of Business
A multiline U.S. property and casualty insurer conducted a 45-day OwlVision POC across five lines of business. See how claims professionals evaluated AI-assisted claim review, Large Loss Report (LLR) generation, line-specific insights, and OwlAssist, with OwlVision generating approximately 70–80% of required LLR content and delivering estimated time savings across the claims tested.
Multiline U.S. Property & Casualty Insurer
Line of businessAuto, General Liability, Property, Professional, and Surety
ProductOwlVision, OwlAssist
Duration45 days
How a multiline U.S. property and casualty insurer evaluated OwlVision across Auto, General Liability, Property, Professional, and Surety claims.
To evaluate how AI could support these workflows, a multiline U.S. property and casualty insurer conducted a 45-day proof of concept (POC) with OwlVision across five lines of business: Auto, General Liability, Property, Professional, and Surety.
The Challenge
Claims professionals often need to review extensive documentation, reconcile information across multiple sources, and translate their findings into structured reports. The time required can vary significantly depending on the line of business, claim complexity, and the adjuster’s familiarity with the file. Preparing a Large Loss Report by hand typically took three to six hours in General Liability, or seven to ten for complex cases, and six to eight hours in Surety.
The Approach
Claim files loaded into OwlVision, with new documentation ingested as it arrived to update the summary, timeline and templates.
Line-specific summaries and timelines, with each insight cited back to the source documents.
Draft LLRs compared against the time reviewers normally spend preparing them by hand.
Conversational questions asked directly against the claim file.
Timeline
- July 2026Pilot begins
A 45-day proof of concept starts across five lines of business.
- Step 2Claims loaded
Approximately two to three claims from each line are included.
- Step 3Feedback sessions
Owl works with claims professionals from each line to review results and gather input.
- Step 4Enhancements
Enhancements are demonstrated and outputs adapted to each line's requirements.
- August 2026Final feedback
Final sessions evaluate OwlVision against existing claim review processes.
Results
| Area | Before | With Owl |
|---|---|---|
| General Liability | 3–6 hours per report; 7–10 for complex cases | About 4 hours saved per report |
| Surety | 6–8 hours per report | About 50% efficiency gain |
| Property | About 1 hour | About 50% saved, to roughly 30 minutes |
| Auto | — | About 25% efficiency gain |
| Professional | — | 1–2 hours saved on familiar files; as much as 7–10 hours on inherited or unfamiliar files |
Time savings for Large Loss Report preparation. Estimates were provided by participating claims professionals during live POC sessions, based on comparisons with their existing manual processes.
What Owl Found
- General Liability
- OwlVision identified contradictory information within claimant accounts, including differences in the reported duration of loss of consciousness.
- Professional
- The platform identified inconsistencies across property documentation where the same property was listed at different values.
- Auto
- OwlVision highlighted differences between the parties’ descriptions of an alleged incident, making the conflicting accounts easier to compare.
- Surety
- The platform provided a breakdown of unpaid subcontractor and supplier claims to help reviewers assess outstanding exposure.
- Property
- OwlVision identified a discrepancy between the occupancy status stated in the policy declarations and information documented elsewhere in the claim.
Reducing the Manual Work Behind Large Loss Reports
Large Loss Report preparation emerged as one of the clearest areas of impact during the POC.
Across the claims tested, OwlVision consistently generated approximately 70–80% of the required LLR content, including key claim information such as claim number, date of loss, and policy information.
The remaining work stayed with the claims professional for review, judgment, editing, and required language.
Reviewers also compared OwlVision-generated LLRs with the time typically required to prepare these reports manually. Their in-session estimates showed different levels of potential time savings depending on the line and complexity of the claim.
Adapting the Workflow to Different Lines of Business
The POC also demonstrated that claim review requirements differ significantly by line of business.
Rather than relying on a single generic claim structure, OwlVision generated line-specific timelines and outputs based on the type of claim being reviewed.
For example, Surety claims included a tailored Bond Timeline, while Professional claims demonstrated the generation of an additional Coverage Analysis Summary Table using DICE Analysis.
As new documentation was added to a claim, OwlVision could ingest the additional information and update the relevant claim summary, timeline, and templates.
This allowed POC participants to evaluate the platform using evolving claim files rather than only static document sets.
Using OwlAssist to Ask Questions Directly Against the Claim File
Participants also tested OwlAssist, Owl’s conversational AI capability, throughout the POC.
As the pilot progressed, users increasingly incorporated OwlAssist into their claim review process, asking specific questions about policy information, claim chronology, reporting timelines, correspondence, and medical information.
Examples of questions asked during the POC included:
- “How much is the liability deductible?”
- “Where can I find the insured vehicle on the policy schedule?”
- “How much time between the date of loss and the date the claim was reported?”
- “How quickly were acknowledgment letters sent out in response to the claims?”
- “What was the first date the claimant complained of neck pain?”
The questions illustrate how users moved beyond general summarization and used OwlAssist to locate specific information within the claim documentation.
In Their Words
“I like the LLR. I think it definitely provides the nuts and bolts... the necessary information that the adjuster would need... the editing is probably gonna be minimal.”
“My testing's been very comfortable with the platform. I like the claim summarization... it gives me as an adjuster kind of an efficient way to understand the current file posture.”
“I thought it was a solid A.”
Key Takeaways
- OwlVision generated approximately 70–80% of required Large Loss Report content for the claims tested.
- Estimated time savings ranged from approximately 25% in Auto to around 50% in Property and Surety, with potentially greater savings on complex or unfamiliar files.
- Timelines and templates were line-specific and updated as new documentation was ingested, with discrepancies cited back to source documents.
- The claims professional kept the final say: reviewing the generated information, applying judgment, editing, and completing required language.