Insurers use software to value your claims. What that means for your demand.

When an adjuster puts a number on a bodily injury claim, software is often part of how they got there. The best-known example is Colossus, and plaintiff lawyers have argued about it for more than twenty years. Here is what that kind of software is, what is and isn't known about it, and what it means for the demand you send.

What claims-evaluation software is

Colossus is licensed to insurers for their claims departments to use when evaluating bodily injury claims. A 1999 report by Connecticut's Office of Legislative Research described it that way, and recorded the vendor's explanation that the software is used to standardize the values assigned to bodily injuries, with the insurer assigning the dollar value of the claim (Connecticut OLR). It can be customized to an insurer's needs, according to the same report.

In practice, the software generates a settlement value from information the adjuster provides about the injury (Minnesota Lawyer). Each carrier decides which tools it uses and how it configures them.

That last point is the one to hold on to. Nobody outside a carrier knows exactly how it has set up its software. Be careful with anyone, us included, who claims to know the formula.

The criticism, and the other side

Plaintiff lawyers have long said evaluation software produces low offers and ties adjusters to ranges they can't go above without a supervisor's approval. In a 2006 Minnesota Lawyer piece, one plaintiff attorney said the offers are "never ridiculously high" (Minnesota Lawyer). The same article quotes the vendor saying the software increases consistency and reduces subjectivity in claims evaluation, and quotes another plaintiff attorney who doubted it had lowered claim values overall.

Regulators have started to address the wider question. In December 2023 the National Association of Insurance Commissioners adopted a model bulletin on insurers' use of AI, which says decisions made or supported by AI systems must comply with all applicable insurance laws and regulations (NAIC).

You don't have to settle that argument to act on it. Whatever you think of the software, it shapes how your demand is read.

The software can only value what's in the file

Whatever tool an adjuster uses, it works from what the adjuster enters, and the adjuster enters what your demand package supports. An injury that isn't documented, a diagnosis that appears only in your cover letter, or a stretch of treatment with no records behind it is hard for anyone to value, by software or by hand.

That makes the ordinary disciplines of a demand matter more:

• Get every record. A provider you didn't request from is treatment the carrier won't see. Track every request until it's fulfilled. Our guide to California medical records request laws and the records request letter generator help.
• Use the records' own words. Diagnoses, findings and restrictions are strongest quoted from the chart, with the page they're on, rather than paraphrased.
• Build a treatment chronology that explains itself. Dates of service, providers, findings and the reason for each next step. If there's a gap in treatment, say why, with the record that shows it. See how to build a medical chronology.
• Get the specials right. In California, recovery for past medical expenses is generally limited to what was paid or incurred, not the full billed amount, under the rule known as Howell. A demand that overstates specials invites the adjuster to discount everything else. Know your liens too.
• Show the effect on the client's life. General damages are where software-driven valuation is most often criticized. Tie the client's limits to what the records and witnesses say, so they're documented, not just asserted.
• Meet the statute when you set a deadline. A time-limited demand in California has to include, among other things, reasonable proof sufficient to support the claim, which may include medical records or bills (Code of Civil Procedure 999.1). See the CCP 999 guide and our policy-limits demand guide.

None of this is new advice. What's new is how much of it a firm can now do quickly, if the work is checked.

Where AI helps on your side

The work above is mostly reading: finding every provider, every date, every diagnosis, every bill, and noticing what's missing. That's where AI on the plaintiff side earns its keep, provided every fact it finds is tied to its page so your team can check it in a click. A chronology with no citations just moves the reading from the AI to the adjuster, or to you.

In PlaintiffStack's personal injury practice, case tools build a missing records report, a treatment chronology and a policy-limits demand letter, and each fact they use cites the document and page it came from. Morning checks flag treatment gaps, record requests nobody has touched, and demands or offers about to expire. Agents propose; a person accepts, and nothing goes to a carrier until someone on your team sends it. Settlement statements and lien reductions sit on the same record as the demand.

One limit matters here. Agents read searchable PDFs, Word and text files, but not scanned records or faxes yet. If a provider sends a scan, a person on your team reads it.

We don't know how any carrier has configured its software, and we won't claim our demands beat it. What we can do is help your team send a demand where every number and every finding can be traced to its source in seconds. See the personal injury practice page, the personal injury case workflow guide, or what Colossus means in our glossary.