ACORD 125 Explained - Every Field, and How to Stop Retyping It

The ACORD 125 is the Commercial Insurance Application, the standardized form that opens a commercial property and casualty submission. It captures the applicant, its premises, its operations, its prior carriers and its loss history, and the line-of-business forms attach behind it.

That is the whole job. Cover page of a commercial account, and everything else in the file hangs off it.

Which would be fine, except nobody reads an ACORD 125 once. You read it into the agency management system. Then into the submission email. Then into a carrier portal at 5pm, PDF on one monitor, AMS on the other, checking the FEIN a third time because you have stopped trusting your own eyes.

So: every field on the form, the forms that attach behind it, and eight questions that separate extraction software which survives a faxed loss run from software that only survives the demo.

Key takeaways

• The ACORD 125 is the account-level application. It says who the applicant is. The forms behind it say what they want covered: the 126 for general liability, the 140 for property, the 127 for business auto, the 131 for umbrella.
• ACORD stands for the Association for Cooperative Operations Research and Development. The forms are standardized in layout, which is not the same as standardized in data.
• The edition stamp in the bottom corner matters. Field positions move between editions, and move again in carrier-modified versions, which is why coordinate-based extraction keeps breaking.
• McKinsey's study of commercial P&C underwriting found that 30 to 40 percent of underwriting's time goes on administrative tasks such as rekeying data. That was 2019. The form has not changed since.
• A completed ACORD 125 can be read automatically and returned as structured fields. The question worth asking a vendor is not the accuracy percentage. It is what the tool does when it is unsure.

Every field on an ACORD 125, not just the section names

Most explanations of this form stop at the section names. Nobody retypes a section name. These are the fields.

• Applicant identity. Legal business name, DBA, mailing address, FEIN, entity type (sole proprietor, partnership, LLC, corporation, joint venture), website, years in business, SIC and NAICS codes.
• Contacts. Named contacts for inspection, accounting and claims, with phone numbers and email addresses, plus the producer and agency details, National Producer Number and state license number.
• Policy information. Proposed effective and expiration dates, billing plan, payment plan, audit frequency, whether the account is direct bill or agency bill, and whether a premium finance agreement is involved.
• Lines of business requested. Which coverages are being applied for, and the premium indicated for each. This is the field that tells the underwriter how many supplemental forms should be sitting behind this one.
• Premises information. Every location: street address, city, state, ZIP, county, whether the applicant owns or leases, interest, full-time and part-time employee counts, annual revenue and occupied area.
• Nature of business. The description of operations, in the applicant's own words. The single most under-written field on the form, and the one underwriters complain about most.
• Additional interests. Mortgagees, loss payees, additional insureds, lienholders, with their addresses and their interest in the property.
• General information. Around fifteen yes or no questions covering prior cancellations and non-renewals, bankruptcies, judgments, uncorrected fire code violations, foreign operations, hazardous materials and safety programs, each with an explanation field when the answer is yes.
• Prior carrier information. Carrier name, policy number, effective dates and premium, usually for the last three years.
• Loss history. Date of occurrence, description, amount paid, amount reserved, and whether the claim is open or closed, typically over three to five years.
• Signature block. The applicant's signature, the date, the producer's signature, and the fraud warning language that varies by state.

Count them. North of a hundred discrete values on a fully completed application, before a single supplemental form is attached. That is what a submission actually is, and it is why intake is the half of this job nobody volunteers for.

How to read an ACORD 125 in one pass

This is the first page of an ACORD 125.

Top down, the logic holds. Producer at the top left, applicant at the top right, policy terms below them, then the coverages requested, then the premises, then everything the underwriter needs to know about the risk that is not specific to any one line: operations, prior insurance, loss history.

Two things trip people up. The audit box, usually marked for an annual audit, which people leave blank without realizing it changes how the premium settles. And the description of operations, where "landscaping" is not an answer and "commercial landscaping and snow removal, no tree work above 15 feet, no herbicide application" is.

ACORD 125 vs 126 vs 140 vs 130

The numbers are the vocabulary of a commercial submission, and nobody explains them to you on your first day. People write it as ACORD 125, ACORD form 125 or just the 125, and they all mean this one. Here is the whole set, in the order you will meet them.

The short version: the ACORD 125 carries the account-level information that applies to the whole submission, and the ACORD 126 carries the general liability details. The 125 answers who the applicant is. The 126 answers what its general liability exposure looks like. Most general liability submissions go out as ACORD 125 and 126 together.

The ACORD 25 is the one that causes real trouble, because it also has a number and it also says insurance on it. It is proof of coverage that already exists. It applies for nothing.

The form gets signed. Then the typing starts.

Nothing on this form is decorative. Every field lands somewhere downstream, which is why a wrong entry on an ACORD 125 is rarely just a wrong entry.

Operations, size and premises drive appetite and class, so a wrong SIC code on the way in is wrong all the way through. Payroll figures, revenue, employee counts and prior premium feed the rating, where one transposed digit becomes a price you cannot honor. Loss history and the general information answers are the underwriter's read on whether this account has a pattern, and they are also the most tedious rows on the form to key. Bad combination.

Then there is the paper trail. The application is a document of record: what was disclosed, when, and by whom. Key it into three systems and you now have three versions of what was disclosed. And none of it stops at binding, because renewals, endorsements and mid-term changes all start from what is on file. If the file is a PDF, someone reads it again.

Manual data entry is not slow because people are slow. It is slow because a hundred fields is a hundred fields, and the second pass is never faster than the first.

Why traditional OCR keeps failing on ACORD forms

The obvious answer is OCR, and the obvious answer has been disappointing people for twenty years. Traditional optical character recognition reads a document into text. It does not read a document into data. On a form like this one, that difference is everything.

Four things break it:

• Edition drift. Coordinate-based extraction learns where a field sits. The 2016/03 edition puts it somewhere, an older edition puts it somewhere else, and a carrier-modified version puts it somewhere else again. Every layout change is a remapping project.
• Checkboxes. A form this dense is mostly boxes, and a box has no text to recognize. An X, a tick, a filled circle and a scrawl all mean yes.
• Handwriting. Plenty of these still arrive completed by hand. Plenty arrive as a scan of a fax of a print of a PDF.
• Tables that run. Loss history and premises schedules break across pages, and a row means nothing without its header.

Vision AI reads the form differently. Instead of looking in a fixed position for a value, it reads the page the way a person does and works out which value answers which question. AI OCR is not better character recognition. It is the difference between getting text out and getting data out.

Eight questions to ask before you automate ACORD 125 extraction

Every vendor in this category will show you a clean digital PDF of a blank form filling itself in. So will the next one. Clean forms extract reliably almost everywhere, which is exactly why that demo tells you nothing.

These are the questions that separate tools. Put them to every name on your shortlist, ours included.

1. Which editions? Ask specifically about 2016/03, older editions still in circulation, and carrier-modified versions. If the answer involves mapping field positions, you will be remapping forever.
2. Which inputs? A typed fillable PDF is the easy case. Ask about flattened PDFs, scans, faxes and handwritten entries, then send them your worst example rather than accepting the answer.
3. Which fields, named? Not "applicant information". Named insured, DBA, FEIN, entity type, premises addresses, SIC and NAICS, description of operations, prior carrier and premium, loss history rows, signature date. If a vendor will not enumerate, that is the answer.
4. What happens on an uncertain field? The single most useful question on this list. A tool that flags low confidence for human review is worth more than a tool with a higher accuracy claim and no way to tell when it is wrong.
5. What comes out? JSON, CSV, Excel, a webhook, a direct API response. Ask to see the actual output shape, not a screenshot of a dashboard.
6. Can it handle the whole packet? Real submissions arrive as a 125 plus a 126 plus a 140 plus loss runs plus a schedule of values, often inside one email. Ask whether the tool splits and classifies the packet or expects you to do it.
7. What does it do about your AMS? Be careful here, and be careful with anyone who answers too quickly. Whether data can be written into Applied Epic, AMS360, Sagitta, HawkSoft, QQCatalyst or EZLynx depends on what your AMS permits and what API access your agency has, not on the extraction vendor's marketing page. Ask your AMS vendor first, then ask the extraction vendor to deliver the shape your AMS accepts.
8. Where does the data live, and is it used for training? An ACORD 125 carries payroll, locations and loss history for a real business. Ask where it is stored, how long it is kept, who on the vendor's side can see it, and whether documents are used to train models. Get it in writing.

Where Parseur fits, and where it does not

Parseur does the extraction layer. It reads a completed ACORD 125 and returns the fields as structured data in JSON, CSV, Excel, a webhook or a direct API response.

It is not an agency management system, not a rater, not a submission platform. If what you need is software that manages the account, you need an AMS, and you can stop reading here with our blessing.

Here is our own answer to the eight questions above.

No templates, no field mapping. The AI engine reads the form and returns the fields you asked for, so there is nothing to remap when the edition changes or a carrier ships its own version of the layout.

Intake is email-native. Forward the submissions inbox to a Parseur mailbox and the attachments get processed as they land, packets included. You name the fields yourself, which means the output arrives in your agency's vocabulary rather than ours, and it leaves as JSON, CSV, Excel, a webhook, a direct API call, or through thousands of applications on the automation platforms. Where your AMS accepts inbound data, Parseur delivers it in that shape. Where it does not, no extraction vendor can fix that for you, and you should be suspicious of one that says otherwise.

On question four, our answer is narrower than what you will hear from some of this list. Parseur does not assign a confidence score to individual fields. What it has instead is an optional validation step: switch it on and every extracted record pauses for a person before it exports, so a wrong value gets caught at intake rather than at renewal. Normalization and validation run on every field either way, holding dates, numbers and choices to the schema you defined. If per-field confidence scoring is the thing your workflow hangs on, better to know that now than in week three.

On question eight, the one about a real business's payroll sitting in someone else's system: Parseur is EU-hosted and GDPR compliant, and customer documents are not used to train AI models. It is written down in the trust center, not buried three weeks into a procurement conversation.

How to extract data from an ACORD 125 with Parseur

No template to build and no field positions to map, so the whole setup is five steps.

1. Create a Parseur account and open a mailbox for commercial submissions.
2. Send an ACORD 125 to that mailbox, or forward the submissions inbox to it.
3. The AI engine processes the file automatically and returns the extracted fields.
4. Name and refine the fields so they match what your agency management system, rater or spreadsheet expects.
5. Send the data onward to a spreadsheet, a webhook, an automation platform, or straight into your own API.

One note, because it is the honest version. Automating extraction does not automate underwriting. Appetite, class and pricing still need your judgment. It only removes the part where the person who knows the account best spends Tuesday morning as a typist.

Test it on the submission you dread

Pick the worst file in your inbox. The scan of a fax, with the handwritten loss history and the description of operations running into the margin. Run that one before the tidy digital PDF, because the tidy one proves nothing.

Then check the boring part, which is question four on the list above. When the tool is unsure about a field, does it say so, or does it fill something in and let you find out at renewal?

You can run that test on your own documents, in your own account, without booking a demo or talking to anybody. The free plan has every feature unlocked, so the ugly fax can go through this afternoon.

If your problem is bigger than one form, the ACORD 125 commercial insurance application page covers this form as a data source, insurance data extraction covers the wider document set including loss runs and declaration pages, claims processing covers what happens after something goes wrong, and insurance automation is the whole picture.

Last updated on August 18th, 2026