A schedule of insurance lands in an inbox at 7:42am. Forty-three buildings, each with a square footage, a year built, a construction class, a sprinkler note, a deductible, an address. The CSR opens it, opens the AMS, opens the carrier portal, and starts typing. Building one, then building two, then building three, the same handful of fields each time. By the time they reach building seventeen, they have been working for ninety minutes and have made three small typos they will not catch until later.
This is what most days look like at a typical independent agency. Most of the data flowing in arrives as something a person has to read and then transcribe. Emails, PDFs, ACORDs, schedules of insurance, loss runs in twelve different formats. The AMS sits on the other side of all of it with empty fields and dropdowns, waiting for someone to populate it. Industry analyst estimates put unstructured data at 80 to 90 percent of new enterprise data, and there is no real reason to think an insurance agency lands on the lower end of that range.
A tax that does not show up on the P&L
Call this the unstructured data tax. It does not appear on any agency's books, but it is paid every day, in CSR hours, in producer hours, and in the silent error rate of records keyed in by humans under time pressure.
The visible portion shows up directly in industry benchmarks. The 2025 PropertyCasualty360/PIA National agent survey puts administrative load at 2.5+ hours per licensed agent per day, most of which is data movement rather than client work. Benchmarks from Inaza's ACORD automation analysis put a single ACORD form at roughly 45 minutes of hands-on prep, and a full submission for a large commercial account at up to 12 hours. Multiply that by the new submissions, renewals, and endorsements your agency processes in a month and you have the visible portion of the bill.
The invisible portion is more expensive. Every time a CSR types "R&B Holdings LLC" into a new account because they could not find the existing "R and B Holdings" record under another producer's book, the duplicate becomes a permanent feature of the database. Industry reporting by Patra puts duplicate records across the AMS and CRM at up to 30% of an average agency's database. Those duplicates are the cumulative residue of the tax, paid in keystrokes over a decade. Cleanup tools can mask the problem temporarily; the architecture keeps producing them.
Why the architecture made sense in 2010
It is worth being fair to the systems being criticized. Applied Epic and AMS360 were architected when the alternative to a human keying data in was nothing usable. OCR was unreliable. Email parsing was a research project. Carriers sent unpredictable formats and were not going to standardize because a software vendor asked. The human-as-ETL pattern was load-bearing because nothing else worked end-to-end. An agency in 2010 paying this tax was not making a mistake. They were paying the going rate for the only architecture that produced reliable records.
This is also why the AMS field types look the way they do. A traditional AMS is an opinionated database with forms in front of it, designed for a person at a keyboard moving at the speed of a forty-words-per-minute typist. Every field is a place where the human is supposed to translate one format into another.
What changed
Three things changed in the last few years. Document parsing crossed the threshold from research project to production technology. Email understanding crossed it shortly after. And the verification problem, the one that used to make extraction unsafe to put in front of a CSR, got solved: a system can now show its work, point at the exact span in the original document where every extracted value came from, and let the human accept or correct it in seconds.
McKinsey's recent work on P&C underwriting describes carriers building underwriting workbenches that do exactly this kind of extraction at scale, with humans focused on judgment instead of transcription. Carriers sit upstream of agencies in the workflow, and they are already moving on this. Most agencies are still typing.
The cost compounds
A tax paid in keystrokes grows with the volume of the work, not with the size of the agency. A 15-person shop that doubles its book size doubles the bill even if it does not double its staff. Industry reporting by Patra on AI automation in agencies puts the recoverable team time at 20 to 30 hours per week across policy checking, quote comparison, and submissions intake. That is the size of the bill being paid right now.
The compounding works the other direction too. Every duplicate record makes the next renewal harder. Every typo in a building address creates a potential E&O exposure two renewal cycles from now. Every email filed against the wrong account creates a small lookup cost the next time someone goes searching. The bill grows with every cycle the architecture stays in place.
What it looks like when the tax goes away
A schedule of insurance lands in an inbox at 7:42am. The system reads it, recognizes it as a schedule for an existing account, extracts the forty-three buildings with their square footage, year built, construction class, sprinkler notes, and addresses, and queues a draft update for the CSR's review. When the CSR opens the work, the comparison view is already open. Two of the buildings are new. One has a different address than what is on file. The CSR confirms three things and approves the rest.
The ninety minutes of typing become ten minutes of looking. Three potential typos do not happen because the data was never typed.
This is what it looks like to stop paying the tax. The change is structural rather than cosmetic. The system reads the documents and proposes the answers; the human reviews, corrects what is wrong, and approves the rest. A chatbot in the corner of the screen cannot produce this outcome, because the chatbot is not where the leverage is.
The tax is invisible because everyone pays it. That does not mean it is free.
We built Aluna so the system reads the documents and the staff review the decisions. Book a demo and we will walk you through a real schedule of insurance, ingested live on the call.