Omnilex

For P&C, health & workers' comp

When a claim is challenged,
the file is the case.

Claims decisions carry bad-faith and regulatory exposure under state law. That exposure turns almost entirely on whether a decision was reasonable and well documented at the time it was made, not just whether it was ultimately correct. In discovery, what gets produced is the whole case: what was reviewed, what was not, and whether the reasoning was consistent with similar claims.

Omnilex is a system and a legal team that reads every claim file end-to-end before the decision is finalized. The system reads the file in full, structures the timeline and findings, and flags contradictions. A lawyer reviews and signs the recommendation. The system never decides a claim; lawyers do. The record exists before litigation, not reconstructed for it.

How it works

From claim file to a record built at decision time

Follow how a claim file moves through the review: from the moment it arrives, on any line, to a documented, lawyer-signed recommendation in the file.

01 · The exposure

Reasonable, and documented at the time

Bad-faith exposure does not turn on whether a decision was ultimately correct. It turns on whether the decision was reasonable, and whether the file shows that it was, at the moment it was made. A sound decision with a thin file carries almost the same risk as an unsound one.

At scale, quality assurance samples a fraction of the book. The files outside the sample carry whatever record they happen to have.

02 · The discovery test

The whole case, produced

When a decision is challenged, discovery does not stop at the decision letter. What gets produced is the whole case: what was reviewed, what was not, and whether the reasoning was consistent with how similar claims were handled.

Bodily injury & liability files

Long files, many parties, many reports. More room for the record and the decision to drift apart.

Health claim determinations

High volume, decided quickly, closely regulated. Exactly the files a sampling regime rarely reaches.

Workers' comp causation

Causation and medical findings carry the file. The reasoning has to hold up years after it was written.

Individual disability files

Long-running and medically complex. Every decision along the way becomes part of the record.

03 · How the review works

Every claim file read end-to-end. A lawyer signs.

! P&C · Health · WC
Step 1 / 6

A claim file arrives

Any line, any length. A liability file of forty pages or an individual disability file running to hundreds.

OMNILEX Every page TimelineFindingsTerms
Step 2 / 6

Every page is read

The system reads the file end-to-end, not a sample of it. The timeline is structured and the findings are set against the policy terms.

Dates that conflictReport vs statementOut of step with the book...
Step 3 / 6

Contradictions are flagged

Anything that does not line up is surfaced: dates that conflict, a report that contradicts a statement, a decision out of step with similar claims on the same book.

Reasoned recommendation
Step 4 / 6

The recommendation is drafted

A reasoned recommendation, consistent with how similar claims were handled, written to be read later.

A named lawyer
Step 5 / 6

A lawyer reviews and signs

Nothing returns as machine output. A lawyer works through the file, the flags and the draft, and signs the recommendation.

Read → Reasoned → Signed
Step 6 / 6

The documented recommendation returns to your team

Signed, documented, and ready for the record. Your team finalizes the decision with the reasoning already in the file.

04 · What you get

A record that exists before litigation

Every claim decision is finalized with a lawyer-signed, documented recommendation already in the file. When a decision is challenged, the record is produced as it was built at decision time; nothing has to be reconstructed. Full review also finds what sampling cannot. For one of our clients, moving from sampled to full review surfaced 2% of decisions that had been wrongly approved, summing up to over $11M a year.

Every claim fileread end-to-end, not sampled
At decision timea record that exists before anyone asks
2%of decisions surfaced as wrongly approved on one book, worth $11M+ a year
Get in touch