From Three Weeks to Minutes: What Claims Automation Is Actually Delivering

Will Drewes

Founder, Fern Strategy · 3 min read

The gap between pilot and production is closing

Insurance claims have always been the operational stress test: high volume, regulatory exposure, and customers who are already unhappy by the time they file. For years, automation in this space was mostly aspirational. The numbers coming out now suggest that's changing, and the results are specific enough to be useful.

Here is what measured deployments are actually showing.

Lemonade: 2 seconds, ~50% straight-through processing

Lemonade approves and pays certain property claims in 2 seconds, with roughly half of all claims now handled by AI without any human intervention. That is not a cherry-picked edge case. It reflects a claims architecture built around straight-through processing from the start, where the AI reviews, adjudicates, and triggers payment in a single pass.

The number matters beyond the headline. At 50% automation, Lemonade's human adjusters are handling the other half, which means complex, contested, or ambiguous claims get more attention, not less. Speed and triage quality move together.

A 400,000-claim travel insurer: 0% to 57% automation, three weeks to minutes

A large US-based travel insurer processing 400,000 claims per year started from zero automation. After deployment, 57% of claims moved to automated processing, and average handling time dropped from three weeks to minutes for those claims.

The scale here is what makes it credible. At 400,000 claims annually, a 57% automation rate means roughly 228,000 claims per year no longer require manual intake, review, and routing. That is not a productivity improvement. It is a structural change in how the operation runs. The remaining 43% still flow to adjusters, but the queue is fundamentally different.

Industry pattern: 50 to 75% cycle time reduction

Across multiple major insurers, reported cycle time reductions for AI-handled claims are running 50 to 75%, with settlement times dropping from the traditional three-week window to a matter of minutes for eligible claims. This range is consistent enough across carriers that it is starting to look like a floor, not an outlier.

The implication for carriers still running manual-first workflows: the competitive gap on customer experience is widening every quarter.

Frende: Hundreds of staff hours per month recovered

Not every win is measured in seconds. Frende, an insurance company, automated AI-generated email summaries for claims-related vendor communication and recovered hundreds of staff hours per month while also accelerating vendor payments.

This is the less dramatic but often more replicable result. Email handling, status updates, and vendor coordination are not glamorous, but they consume real adjuster time. Automating the summarization and routing layer does not require rebuilding the core claims system. It layers onto existing workflows, which means faster deployment and a cleaner audit trail.

What this means for insurance operations leaders

These results share a common structure. They are not black-box deployments where the AI makes opaque decisions and someone hopes for the best. The carriers seeing the strongest outcomes built automation around defined decision rules, clear escalation paths, and the ability to explain every adjudication. That is not just good practice. In a regulated environment, it is the only design that survives a market conduct exam or a bad-faith claim.

The build-vs-buy question in claims automation is real, but it is secondary to a more important one: can you audit what the system decided, and why? Carriers that answer yes first tend to scale faster and with fewer regulatory surprises.

The two-week diagnostic Fern runs is designed to answer exactly that question before a carrier commits to a direction. If you are evaluating claims automation and want to know what your current workflow can actually support, that is the right starting point.