Short answer: To stop re-keying commercial insurance submissions, capture each account's data once in a structured "master submission" built from the ACORD applications, schedules, loss runs and emails you already collect, then let automation fill each carrier's application, upload or API from that record. AI reads the messy documents and maps your data to each carrier's questions; rules and a person check the fields that drive rating before anything is sent. Done well, a CSR reviews and approves a submission instead of typing it in four times.
The problem
Placing a commercial account means sending largely the same information to several markets. In practice, that often means typing it out again for each one. The 2026 Insurance Agency-Carrier Connectivity Trends survey from Applied Systems, which asked 702 independent agents in April and May 2026, found that re-keying risk data into multiple carrier portals was the top pain point, named by 74% of respondents. Agencies that lean towards commercial lines re-key more often than personal lines shops (70% against 62%).
The workaround is familiar to anyone who has worked a commercial desk. A Celent report commissioned by Ivans describes it plainly: for complex accounts with many buildings, vehicles and schedules, data entry into each carrier's portal is so tedious that most CSRs won't do it. Instead they generate an ACORD application from the agency management system, fill in one carrier's supplemental, attach loss runs and financials, bundle the lot into a PDF and email it. The report notes that this simply moves the typing to the underwriter's side, and the back-and-forth questions follow.
The same report points out a quieter cost on small accounts. When a policy carries a small premium, it is hard to justify entering the same data into several portals, so the account gets quoted with fewer markets than the client deserves. Where two carriers are a tie, the one with the easier portal tends to win.
What it costs
The survey shows that carriers feel this too. In the same Applied Systems research, 90% of agents said they had reduced business with a carrier because of submission friction, and 76% said the digital submission and servicing experience matters more to them than commission.
For the agency, the cost is staff time and slower quotes. Nobody publishes a neutral benchmark for minutes per submission, and it varies hugely with the line and the portal, so measure your own (there is a simple way to do that in the checklist below). Errors are the other cost. A wrong building year, payroll figure or vehicle garaging address typed into one portal and not another means quotes that are not comparable, or a quote that changes at bind.
Why it keeps happening
Every carrier asks the same things differently
ACORD, the not-for-profit standards body set up in 1970, publishes a large library of standard forms and data standards that are widely used across US insurance, according to Wikipedia's summary. The ACORD 125 and its line-specific companions should mean you collect data once. But each carrier also has its own supplemental questions, its own portal screens and its own idea of which fields are mandatory. Some accept an uploaded ACORD PDF, some want the portal filled in, and a few take data directly from your agency management system.
The source documents are unstructured
The information arrives as a filled PDF from last year, a spreadsheet of vehicles from the client's fleet manager, a statement of values in the client's own format, and loss runs from the prior carrier that look nothing like anyone else's. Before anything can be submitted, a person has to read all of that and turn it into answers.
Loss runs add a wait in the middle
Underwriters want the account's loss history from prior carriers. Some states set rules for how quickly insurers must provide it. Oregon, for example, requires insurers or their appointed producers of record to make loss runs available to commercial policyholders within 15 calendar days of a request, covering five years or the whole period insured if shorter. When they arrive, someone still has to summarise them into a format the next underwriter can read. In the 2026 survey, 44% of agents named automated loss run delivery as a capability they want from carriers.
Comparative raters only cover part of the book
Commercial raters and carrier bridges help with small, simple risks such as BOPs, and they are worth using where they exist. Middle market accounts with long schedules, unusual exposures and supplementals still tend to fall back to PDF and email.
How AI solves it
The pattern that works keeps the data in one place and treats every carrier submission as an output of that record. AI is useful in two specific spots: reading documents that arrive in any layout, and matching your answers to each carrier's differently worded questions. Everything else is ordinary workflow automation.
The workflow, step by step
- When an account goes to market, a workflow (n8n, Zapier, Power Automate or a small custom service) gathers the documents into one folder: the client's completed questionnaire, last year's applications, schedules, statements of values, loss runs and relevant emails.
- A document model pulls fields and tables from each file into a structured record. Filled ACORD PDFs often have form fields you can read directly with no AI at all. Scans, spreadsheets in odd layouts and loss runs go to a document or vision model.
- The extracted data is merged into one record per account: named insured, locations, building details, payroll and revenue by class, vehicles, drivers, prior carriers and loss history. Where two documents disagree (a different square footage in the SOV and last year's application), the record flags the conflict instead of picking one.
- A language model turns each carrier's loss run into one standard table by year: claim count, paid, reserved, open or closed, and a short description of anything large. Totals are then recalculated by code, so the summary always adds up to the source.
- For each target market, a mapping layer fills that carrier's format. Where a carrier takes ACORD forms or a direct upload, rules handle it. Where it has its own supplemental, a language model proposes answers from the master record and marks any question it cannot answer from the data you hold.
- A CSR sees one screen per carrier: the filled application, the fields flagged as low confidence or conflicting, and the open questions to send to the client. Nothing goes out without that approval.
- Approved submissions go out by API, upload or email with the attachments in the order that underwriter prefers. The workflow logs what was sent where and chases quotes that have gone quiet.
Where to use AI and where to use rules
| Task | Best tool | Why |
|---|---|---|
| Reading filled ACORD PDFs with form fields | Rules | The data is already in named fields |
| Reading scans, client spreadsheets and SOVs | Document or vision model | Layouts vary from client to client |
| Summarising loss runs from different carriers | Language model, with code checking the totals | Every carrier's format is different; the maths must be exact |
| Matching your data to a carrier's supplemental questions | Language model, with saved mappings | Wording differs; confirmed mappings get reused |
| Validating values such as VINs, postcodes or class codes | Rules and lookups | Exact checks are cheaper and more reliable |
| Deciding which markets to approach | Person, helped by appetite data | A relationship and judgement call |
| Sending the submission | Person approves, automation sends | You are accountable for what the carrier receives |
On confidence: document models typically return a score for each field. Microsoft's guidance for its Document Intelligence service explains that a score of 0.95 means a value is likely correct about 19 times out of 20, and suggests using the score to decide whether to accept a value automatically or send it for human review. Set the threshold high for anything that drives rating, such as payroll, revenue, total insured values and driver details.
A note on carrier portals. Some tools "drive" carrier websites with bots to fill the forms for you. That can work, but it breaks whenever a portal changes, and some carriers' terms may restrict automated access. Prefer each carrier's API, bridge or upload route where one exists, and check the terms before automating a login.
Costs and timeline
Every figure in this section is an estimate based on the assumptions shown. Swap in your own numbers.
Assumptions: an agency sends 60 commercial submissions a month, each to an average of four markets. Re-keying into each carrier's format takes 30 minutes, and a CSR costs $32 an hour fully loaded.
That is 240 carrier submissions and about 120 hours of typing a month, or roughly $3,840 a month in staff time. If automation cut the time per carrier to 10 minutes of review, the agency would spend about 40 hours, freeing around 80 hours a month for client work, follow-up and quoting more markets per account.
| Option | What it involves | Best for |
|---|---|---|
| Do nothing | Keep emailing PDFs and re-keying the portals that require it. No new spend, but the hours and the narrow marketing remain | Very low commercial volume, or a book dominated by one or two markets |
| Use your AMS and rater features | Turn on commercial rating, carrier bridges and ACORD prefill in the systems you already pay for | Small commercial accounts with carriers that integrate with your AMS |
| Buy submission software | A dedicated submission intake or portal automation product, usually priced per user or per submission | Agencies that want a packaged product and whose carriers it already supports |
| Custom or hybrid build | A workflow around your AMS, document store and email, with AI for extraction, loss run summaries and supplemental mapping. Model usage at 60 accounts a month is likely to be tens of dollars, assuming a few cents per page | Middle market books, specialist programmes, unusual supplementals, or a wish to own the data and mappings |
Timeline (estimate): a first version covering document intake, the master submission record, loss run summaries and mapping for your top three markets usually takes six to ten weeks. Most of that time goes on agreeing the master data model and testing extraction on your own past submissions. Run it in parallel with your current process for a few weeks before relying on it.
When not to use AI
- Most of your commercial business is small and simple and your carriers already take data from your AMS. Configure those integrations first; they cost less than anything new.
- A packaged submission tool already supports your main markets. If it covers 80% of your volume, buy it and keep the edge cases manual.
- Your client data is thin. If you do not collect payroll by class, building details or driver lists in the first place, automation cannot fill the gaps. Fix the client questionnaire first.
- The account needs a narrative. A complex risk with a poor loss history needs a producer's explanation to the underwriter. AI can draft a cover note from the facts, but the argument is yours.
- You would be tempted to let it submit unattended. The agency answers to the carrier and the client for what it submits, so keep a person approving every submission.
How to start
- [ ] Time ten recent commercial submissions from start to finish, per carrier, and note where the minutes went.
- [ ] List your top five commercial markets and how each prefers to receive a submission: portal, ACORD upload, email or API.
- [ ] Collect the supplemental questionnaires for those markets and mark the questions that overlap.
- [ ] Gather 30 past accounts with their source documents and the final submissions, to use as a test set.
- [ ] Decide which fields always need a human check before sending.
- [ ] Check what your AMS and rater can already do before building anything.
If you want help building this, our AI workflow automation service covers intelligent document processing, loss run extraction and the integration with your AMS and inboxes. If you are still weighing a packaged tool against a custom build, AI strategy consulting ends in a working prototype tested on your own submissions. For a related insurance workflow, see how to automate COI tracking for subcontractors and vendors. The same document-to-structured-data approach is behind AI order entry automation for distributors.
FAQ
How do I stop re-keying data into carrier portals?
Build one structured record per account and treat every carrier submission as an output of it. Use your AMS bridges and ACORD uploads where carriers accept them, and use automation to fill the remaining formats from the same record. A person reviews each submission before it is sent.
Can AI fill out ACORD forms?
Yes, AI can read client documents and prior applications and propose values for ACORD 125, 126, 140 and similar forms. Fields that already exist in your AMS should be filled by plain integration, not AI. Treat AI output as a draft: check payroll, revenue, values and driver details before sending.
Can AI summarise insurance loss runs?
A language model can turn loss runs from different carriers into one standard table with claim counts, paid and reserved amounts and open claims by year. Have code recalculate the totals from the extracted lines, and compare against the source on large or open claims. That makes the summary quicker to produce and safer to rely on.
Is commercial submission software worth it for a small agency?
It depends on volume and how many markets you approach per account. If you place a handful of commercial accounts a month with one or two carriers, your AMS features are probably enough. Once you are sending dozens of submissions a month to several markets each, the time saved usually justifies a tool or a custom workflow.
Is it safe to put client data into AI tools?
It can be, if you use business accounts or APIs that do not train on your data, keep data in a region you are comfortable with, and limit access to the people who need it. Client financials, payroll and driver records are sensitive, so check your carrier agreements and privacy obligations before you start.
Sources
- Applied Systems press release, 2026 Insurance Agency-Carrier Connectivity Trends survey (26 August 2026): https://www.ivans.com/news/press-releases/2026/agents-are-choosing-carriers-that-automate-submissions-say-findings-in-2026-insurance-agency-carrier-connectivity-trends-survey-report/
- Insurance Business, "Submission friction now beats price in where agents place business, Ivans finds" (27 August 2026): https://www.insurancebusinessmag.com/us/news/technology/submission-friction-now-beats-price-in-where-agents-place-business-ivans-finds-587645.aspx
- Celent, "Commercial Lines New Business Submission: New Tools for a New Day", commissioned by Ivans (September 2020): https://www.ivans.com/globalassets/all-documents/resources/reports/ivans-commercial-lines-new-business-submission-celent-report.pdf
- Insurance Journal (EZLynx sponsored blog), "How Small Agencies Can Win in Commercial Lines Quoting" (14 August 2026): https://www.insurancejournal.com/blogs/ezlynx/2026/08/14/881127.htm
- Or. Admin. Code § 836-080-0810, Provision of Commercial Loss Runs (Cornell Legal Information Institute): https://www.law.cornell.edu/regulations/oregon/Or-Admin-Code-SS-836-080-0810
- Microsoft Learn, "Interpret and improve model accuracy and confidence scores" (Azure Document Intelligence): https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/concept/accuracy-confidence
- Wikipedia, ACORD: https://en.wikipedia.org/wiki/ACORD
If you would like to talk through your own submission process and whether automating it is worth it, get in touch at https://sintraflow.com/contact.
