As Head of U.S. at Upstage AI, Kasey Roh works with insurance teams evaluating AI underwriting automation, document extraction, and submission intake workflows.

If you work in commercial insurance underwriting, your morning probably looks something like this:

A broker sends over an insurance submission package. It's a single PDF, maybe 50 pages, maybe more. Somewhere inside that file is an ACORD application. ACORD applications are standardized insurance forms used to collect and share policy information. Along with that, there may be a few years of loss run reports from different carriers, a supplemental questionnaire, and maybe some financial statements or property schedules. The formats are all different. The layouts don't match. Some pages are scanned. Some are digitally generated. Some are barely legible.Before you can even begin evaluating the risk, somebody has to sort through the whole package, figure out what's what, and pull out the key fields: the insured's name and address, the policy period, claims history by year, and NAICS codes, which identify a business's primary industry or activity. All of that underwriting data gets entered into your underwriting workbench or policy admin system. An underwriting workbench is the digital workspace underwriters use to manage risk review, quote preparation, and related underwriting tasks.

That manual work adds up fast. Accenture's recent underwriting research found that more than a third of an underwriter's time is still spent on non-core work, including data collection and administrative activities. Even a clean submission can take 15 to 20 minutes to sort, review, and enter before actual underwriting work begins. Complex cases take longer.

Meanwhile, the broker is waiting for a quote. If your team is handling dozens of submissions a day, you're spending entire workdays on manual data entry instead of actual underwriting work: evaluating risk, pricing it correctly, and getting back to brokers fast enough to win the business.

The worst part: it's not even hard work. It's tedious work. Your underwriters know how to read these documents. The problem is that they're spending the majority of their time doing something AI-powered underwriting automation should be doing for them.

What Studio does with a submission package

Upstage Studio automates underwriting submission intake by using prebuilt AI agents to classify, extract, and structure data from your broker submission packages. No coding required. No heavy system implementation. Here's what happens:

Upload the documents. You drop the full insurance submission package into Studio, including the PDF, mixed document types, and all.

Studio classifies the documents automatically. This is the step that saves the most time upfront. Studio reads through the entire package and identifies which pages belong to which insurance document type. ACORD application pages get grouped together. Loss run pages get grouped separately. Each supplemental form and financial statement gets classified into its own category.

Why does this matter? Because different document types need different data extraction. From a loss run, you want claims history by year. From an ACORD form, you want the applicant's information, their codes, and their coverage requests. Studio knows the difference and applies the right extraction schema to each document type automatically.

The schema is customizable. The prebuilt extraction works out of the box for common insurance document types. But if your operation has different lines of business or you need different fields, updating the schema is straightforward. You can add or remove keys and adjust field descriptions in plain language to fine-tune how Studio interprets your documents.

Studio extracts structured underwriting data from each document. Once classified, Studio pulls the relevant fields from each document type. The output is clean, structured data, not a wall of raw text. Policyholder name, address, SIC/NAICS codes, effective dates, prior carrier information, claims counts, and incurred amounts by year. All of it is mapped to specific fields and ready to use.

Export to your systems. The structured output can be pushed to your underwriting workbench or policy admin system via API, or exported as JSON. The API specs are available directly in Studio, so your team can grab what they need and start integrating right away.

The entire process, upload, classify, and extract, runs in under three minutes. What used to be 15 to 20 minutes of manual sorting and data entry becomes a quick review of the structured output.

Why underwriting automation matters for your team

The time savings alone are significant. Tricura Insurance Group, an MGU focused on senior living liability coverage, integrated Upstage into their underwriting intake process. The result: roughly 30 minutes saved per single-facility application, plus about 20 minutes for each additional facility. That freed up their underwriting team to handle growing submission volume without adding headcount, while reallocating time toward higher-impact work across the organization.

But the real impact goes beyond efficiency. When your team isn't buried in data entry, they respond to brokers faster. Faster responses mean more quotes out the door, creating more opportunities to bind business. Complex cases that used to get pushed aside because someone was still manually sorting through a routine submission now get the attention they deserve.

And because Studio structures the data consistently every time, you get cleaner underwriting data flowing into your downstream systems. Fewer typos, fewer missed fields, and fewer inconsistencies that cause problems later in the policy lifecycle.

‍Ready to see underwriting submission intake automation on your own documents?

Try the interactive demo to see how Studio classifies, extracts, and structures data from broker submission packages.

Questions? Reach out to our team.