The Timeline Is the Claim
Issue 4 /
From the CEO's desk
A claim is a story that happened in order: someone got hurt, was treated, was billed, made a demand. The documents containing that information do not often come in any sequence. The adjuster's job is to put those moments back in sequence and see whether they hold together. The facts are scattered — a date in a police report, another in a chart, another on a bill — across dozens of documents that arrive in no order at all.
This edition is about the AI we built to reconstruct that order, and to prove it.
A wrong date reads as fact and quietly bends the whole story. So we hold the AI to a simple standard: show the evidence, not just the answer.
The result is two timelines, because a claim has two stories. The Event Timeline tells what happened: every event is marked before, on, or after the incident and links back to the page that proves it. It works for a single document and the whole claim file. The Document Timeline tells how the file itself came together: when each document was dated, created, and added. Order also exposes what is missing: once the treatment events line up, a gap in care shows itself, something nearly invisible when documents are read one at a time.
Customer SpotlightReconstructing a Claim Timeline: Casualty Claim Evaluation Challenge: Dozens of documents per claim, and no single ordered view of what happened when. Building the sequence by hand ate hours, and quiet gaps in treatment were easy to miss. Solution: ClaimLens™ reads every document and builds two timelines: an Event Timeline of what happened and a Document Timeline of how the file came together, every entry cited to its source page. Outcome:
Business Impact: Faster case theory, fewer missed gaps, a clearer basis for valuing the claim. |
DocLens.ai Product Corner
Feature Spotlight: Event Timeline & Document Timeline
Unlike a plain document reader, ClaimLens™ gives you two views:
Event Timeline: what happened in the claim, in order
Document Timeline: when each document was dated, created, and added to the file
And for both:
Reads every document in the claim
Sorts events as before, on, or after the incident
Attaches the exact source page to each event as a citation
Works at two levels: a single document (event timeline only), or the whole claim file
Flags gaps in treatment where the record goes quiet
Keeps guesswork out: every entry is backed by evidence
AI This MonthWhen the Evidence Itself Is Fake For years the worry about generative AI was the text it writes. The newer worry is the evidence it fakes. It is now trivial to produce a photo of damage that never happened or a document that was never issued, and those files are beginning to land inside real claim files. A claims platform that cannot tell a real page from a generated one is a platform that can be fooled. So we added detection that flags AI-generated images, and PDFs that contain them, before they reach a reviewer. As the tools to fake evidence get better, the tools to catch it have to keep pace. |
Knowledge Corner
The 2,000-Page Problem
A real document pushed us to fix something quiet but important. One medical record ran 2,119 pages and 186 megabytes. Our splitter was meant to cut it into a handful of workable files. Instead it produced close to a thousand, each one still carrying the full weight of the original. The cause was small and stubborn: image metadata was being copied into every page, so each slice was still measured as if it held all the pictures in the book. The fix was to strip that metadata before measuring size. The lesson travels: at scale, the thing that breaks you is rarely the logic, it is the hidden weight you forgot you were carrying.
Also Shipped
Portfolio Analytics in plain English. Ask your whole book of claims a question the way you would ask a colleague, and get an answer back — no query language required.
Quality Audit, still improving. We kept sharpening the automated quality audit we introduced last edition, so the result stays in step with a reviewer's own overrides.
Subscribe





