Straight answers
How do you automate loss run processing?
Loss runs are the carrier's official claim-history report, and they arrive in every format, so automating them is really a document-extraction problem: read the file, whether a clean PDF, a scan, or a spreadsheet, and turn each claim into structured fields like date of loss, amount paid, amount reserved, and open or closed status. Tools built for loss runs include Sensible, Docugami, and Deep Vector's Loss Scan, alongside broader insurance document platforms like Convr and Docsumo. For an independent agency, Relay's Document Parsing is live: it reads loss run PDFs into structured claim data and flags any field it is unsure about for a person to verify. The goal is to stop typing claim histories into a spreadsheet by hand and check the extraction instead.
| Stage | By hand | With extraction |
|---|---|---|
| Get the loss runs | Request from each carrier, usually the agent portal, sometimes email to underwriting | Still a request step; automation begins once the file is in hand |
| Read the document | Squint at a PDF or scan, one claim at a time | Extraction reads the file and returns claim-level fields |
| Structure the claims | Type date, paid, reserved, and status into a spreadsheet | Fields come out structured: date of loss, paid, reserved, open or closed |
| Catch bad reads | Hope nobody dropped a claim or mistyped an amount | Uncertain fields flagged for review (Relay Document Parsing, live) |
The longer version
Loss runs are hard to automate because no two carriers format them the same way. One sends a tidy PDF, the next a scan of a printout, a third an Excel export with merged cells. A tool that nails one carrier can choke on another, so test extraction on the carriers you actually pull from. Underwriters usually want three to five years, and workers comp submissions often ask for five, so a single account can mean several documents in several formats.
What processing should produce is claim-level structured data your team can total, sort, and drop into a submission or a spreadsheet: date of loss, cause, amount paid, amount reserved, and status. Relay's Document Parsing reads loss run PDFs into that structure and attaches a confidence check to each field, marking shaky reads for review. Parsed claims land in the same client record as the dec page and the ACORDs, so the whole account sits in one place.
One honest limit: extraction handles the reading and the typing. It does not judge the account. A licensed person still reads the loss history, decides what it means for appetite and pricing, and whether to remarket. Automation removes the transcription, not the underwriting.
Common questions
What data comes off a loss run?
For each claim: date of loss, claim type or cause, amount paid, amount reserved, and whether it is open or closed, plus the policy period. Extraction turns that into structured fields so you can total and sort it instead of reading down a PDF.
Can Relay read loss runs?
Yes. Document Parsing is live and reads loss run PDFs into structured claim data, with a confidence check on every field. Anything it is unsure about is flagged for a person to verify. It does not judge the account; a licensed person still makes the underwriting call.
Related questions
Part of the Relay straight answers library. Updated 2026-07-11. See how we source content.
Get started
Get on the Summer ’26 list.
Spring ’26 is full. Get on the Summer ’26 list and we will be in touch as spots open.