The Omaha Agency That Stopped Retyping

This is a composite scenario — not a named client, but built from the real patterns that show up in commercial insurance agencies across the Omaha market. The numbers, the roles, and the friction points are representative of what agencies working in the Midwest commercial lines space actually face. If it sounds familiar, that’s intentional.

The agency in this walkthrough is mid-sized: 80 employees, a commercial lines book that spans contractors, manufacturers, and agribusiness accounts, and a team of six underwriting coordinators who handle submission intake. The agency runs a leading AMS and has been on it for years. The problem isn’t the system. The problem is what happens between the broker’s email and the system.

The before — a senior underwriter retyping 30 submissions a week

The intake workflow starts the same way every day: a batch of broker emails arrives with attachments. ACORD 125 applications, supplemental questionnaires, sometimes a Word document with loss run data pasted in, sometimes a spreadsheet. The coordinator opens the email, opens the AMS, and starts typing. Named insured, SIC code, prior carrier, five-year loss run totals, coverage requests, limits.

A straightforward commercial package application takes about 40 minutes from email to AMS entry. A complex contractor submission with multiple layers of coverage and a loss run narrative can run to 90 minutes. The coordinators are doing 30 submissions a week combined. That’s 20 to 30 hours of data entry — skilled people doing work that requires no judgment, just accuracy and patience.

The senior underwriters are waiting for the queue. When intake is behind — which it is every time the agency runs a marketing campaign or a carrier pulls out of a line and triggers a remarket rush — underwriting waits. Submissions age in the inbox. Brokers follow up. The intake log gets longer before it gets shorter.

Why generic AI tools don’t solve this

The coordinator’s first experiment with generic AI tools starts the way most AI experiments in insurance agencies start: someone tries to run a few submissions through a general-purpose AI assistant, ask it to extract the key fields, and paste the output into the AMS.

It half-works. For clean, standard ACORD forms with typed entries, the extraction is reasonably accurate. For older broker PDFs with handwriting, marginal notation, inconsistent field labeling, or non-standard supplementals, the output requires so much cleanup that it isn’t faster than typing from scratch. More importantly, the extracted data isn’t in the AMS — it’s in a chat window, and someone still has to key it in.

Generic tools don’t solve the problem because the problem isn’t just extraction. It’s extraction plus validation plus AMS population plus routing. A general-purpose AI has no connection to the AMS data model, no understanding of the agency’s carrier appetite, and no way to flag a submission for the right underwriter based on line of business and account size. Building those connections is the work.

What an in-production submission-intake build looks like (AMS, ACORD parsing, severity tagging)

The production intake pipeline has a few moving parts, all integrated.

An email-monitoring service watches the submissions inbox. When a new email arrives with attachments, it routes the document set into the intake pipeline. The pipeline runs ACORD-aware extraction — trained to handle the variation in commercial lines forms across different carrier and MGA templates, including the supplemental forms that don’t follow ACORD formatting at all.

Extracted fields map to the AMS data model. Named insured, address, business type, SIC code, coverage type, limits, prior carrier, five-year loss totals. The mapping isn’t one-to-one; some AMS fields require inference from multiple source fields, and the pipeline handles that translation. A confidence score goes with each field.

Severity tagging runs alongside extraction: the submission gets a preliminary risk category based on SIC code, loss run profile, and coverage requested. The tag isn’t an underwriting decision — it’s a routing input. High-complexity submissions go to the senior commercial underwriter. Standard accounts go to the standard queue.

The AMS entry is created as a draft. The coordinator gets a review screen: extracted fields on the left, source document on the right. Any field with a confidence score below threshold is highlighted. The coordinator reviews, corrects, and approves. The draft becomes a live submission in the AMS.

The week-3 morning where the queue is empty

Three weeks into production, the intake supervisor opens the queue on a Wednesday morning and finds something she hasn’t seen in months: it’s current. Every submission from the prior day is in the AMS and assigned. She sends a message to the COO that says, basically, the thing is working.

What changed isn’t that the coordinators are doing less. They’re doing different work. Instead of spending 40 minutes entering data from a PDF, they’re spending 8 minutes reviewing a draft that’s already 90% populated and approving it. The difference is that their attention is on exceptions — the handwritten supplementals, the accounts with unusual SIC codes, the loss run narratives that don’t fit the standard parsing template.

Brokers are noticing. Turnaround on submission acknowledgment is faster. The follow-up calls asking “did you get my submission” have dropped off.

Numbers we’d expect to see — hours per submission, throughput, exception rate

These ranges reflect what a well-implemented build typically produces — actual results depend on the agency’s carrier and broker mix, AMS configuration, and how submissions are currently routed before they hit the coordinator queue.

The benchmarks for a well-implemented submission-intake build in a commercial lines agency of this size:

Time per submission: From 40–90 minutes down to 8–15 minutes for coordinator review. The variation depends on submission complexity and the exception rate for that carrier’s form templates.

Throughput: The same team handling 30 submissions per week can handle 80 to 100 with the same headcount, because the pipeline runs continuously — evenings and weekends included. Submissions don’t age in the inbox overnight.

Exception rate: Expect 15–25% of submissions to require meaningful coordinator correction in the first month. That rate drops to 8–12% after the pipeline has been tuned against the agency’s actual carrier and broker mix for 90 days.

Adoption: The coordinators are the ones who benefit most, and they know it. In practice, adoption is the easy part of this build. The part that requires attention is keeping the extraction models current as carriers update their form templates — that’s ongoing maintenance, not a one-time setup.

For more on how Blue Sage approaches submission intake and the other high-value workflows in commercial insurance, see the insurance practice.