Resources / Blog

Field notes on AI agents inside insurance operations.

Long-form pieces on how carriers, MGAs, and health plans are actually deploying AI agents across claims and underwriting. Written by the team at Layerup.

Book a demo Explore platform

Posts

The latest from Layerup.

The most recent piece, opened up. Search and filter the full catalog below.

Platform & engineering

Purpose-built, not general-purpose: why Layerup ships a different agent for every underwriting and claims line

An agent that underwrites IDI is not an agent that underwrites life, and neither is a general-purpose model with a long prompt. The performance gap is architectural, and it shows up in every line-specific detail.

LayerupJune 12, 202613 min read

Browse the catalog.

Search the full library, or narrow it down by category, topic, and audience.

12 posts

Platform & engineering

Purpose-built, not general-purpose: why Layerup ships a different agent for every underwriting and claims line

General-purpose AI demos well on any line and underperforms in production on every line. This is the technical anatomy of Layerup's purpose-built agents — per line, per function — and why specialization wins on the metrics that matter.

June 12, 202613 min read

Claims

Agents that compound: how Layerup's AI improves the more your enterprise uses it

The first agent you deploy is the worst agent you will ever run. This is the engineering behind why Layerup's agents get measurably better on your data — and what that looks like on core claims metrics.

June 4, 202611 min read

Claims

Compressing FNOL-to-payment cycle time from 14 days to 36 hours

The industry talks about cycle time as if it were a property of the claim. It is not. It is a property of the queue. Here is how to drain the queue.

May 22, 202610 min read

Claims

Estimate QA is the highest-leverage AI deployment in auto and property claims

Estimate QA decides what you pay. Reviewing every estimate, line by line, is the single workflow with the largest dollar impact per agent hour.

May 8, 20269 min read

Claims

Closing the subrogation gap: turning recoverable exposure into actual recoveries

Subrogation is not a detection problem. It is a workflow problem. Files with recovery potential get identified, then quietly drop off the radar because the next step is too expensive to take.

April 24, 20269 min read

Claims

Reserve accuracy is the unspoken loss-ratio lever

Reserve drift is rarely a forecasting failure. It is a documentation failure. Files that develop continuously can be reserved continuously.

April 10, 20268 min read

Claims

From signal to substantiated: rebuilding fraud detection around outcomes, not alerts

Alert volume is the wrong metric. SIU referral acceptance rate and substantiated outcome rate are the metrics that decide whether a fraud program is working.

March 27, 20269 min read

Underwriting

Submission throughput is the new growth lever for commercial carriers

Submissions per underwriter is the single number that decides how much premium your underwriting organization can handle. Move it and the rest moves with it.

March 13, 20269 min read

Underwriting

Appetite consistency: how to expand the top line without expanding the risk profile

Carriers do not lose discipline by changing appetite. They lose discipline by applying appetite inconsistently. Consistency is the cheapest growth lever you have.

February 27, 20268 min read

Underwriting

Renewal review at scale: from queue to bound in days, not weeks

Most carriers know exactly how many renewals are coming. Most carriers handle those renewals as if each one were a surprise. That gap is where retention dies.

February 13, 20268 min read

Operating model

From doing to approving: the new operating model for insurance work

Insurance has always run on humans doing the work. The next decade runs on humans approving the work. The operating model has to change with it.

January 30, 202610 min read

Platform & engineering

Deploying AI agents inside modern policy admin and claims cores without ripping them out

Agents do not need a core replacement. They need a write-back contract, a sandbox path, an evals harness, and an audit trail. Here is the architecture pattern that survives a model risk review.

January 16, 202611 min read

Get started

Move work from doing to approving.

Most carriers we work with start with one workflow inside one line of business. Talk to us about where the highest-leverage deployment is in your operation.

Book a demo Explore platform