A bodily injury claim comes in on Monday morning. A routine rear-end collision involving soft-tissue neck and back pain, the kind of minor injury that typically resolves on its own in most files. Nothing about it trips a flag, so it lands in the general queue and waits its turn.
Ninety days later, a lawyer has become involved and the estimated cost of the claim has more than tripled. A single runaway claim like this can quickly wipe out the margin of a hundred well-handled ones.
This is the file that should worry a claims executive, precisely because nothing about it looked like trouble at intake. Expensive bodily injury claims rarely announce themselves on day one. By the time severity is visible in the file, the damage is already done.
This highlights the fundamental blind spot in how most claims operations triage. They sort files by the severity observable at first notice of loss (FNOL), operating on the assumption that ultimate cost can be read off the intake data. The evidence points the other way. What a bodily injury claim finally costs depends far more on the path it takes after FNOL than on how it looks at intake, and that path is partly set by the insurer's own early handling decisions.
The Insurance Research Council has studied attorney involvement in auto injury claims since the 1970s, and the pattern has held: represented bodily injury claimants settle for roughly 3.5 times what unrepresented claimants with comparable injuries receive, and represented claims account for about 85% of all dollars paid out under bodily injury (BI) coverage.
The easy interpretation is that attorneys are drawn to claims that were destined to be expensive anyway. That is partly true, and for years it was hard to separate selection from cause. The more recent work closes that gap. A 2026 study in the Journal of Risk and Insurance, led by Bogdan Savych at the Workers Compensation Research Institute (WCRI), used an instrumental-variables design across more than 950,000 lost-time claims to isolate the effect of representation itself, and found that attorney involvement causally raises indemnity, independent of the underlying injury. The path a claim takes helps determine what it costs, which means some of that cost is still in play at the point an adjuster first touches the file.
The same WCRI research program also looked at why injured workers hire attorneys in the first place, and the answer is rarely the injury. Workers who believed their claim was being denied or delayed retained counsel at around 34%, against roughly 5% among those who sensed no problem. The trigger was the experience of the claims process: whether a claim goes down the expensive path is, to a meaningful degree, a decision the insurer participates in during the first few weeks.
The window for that influence is narrow: a 2025 Milliman analysis that ran natural-language processing across workers’ comp claim notes found that attorneys are identified on only about 8% of claims within two days of first report, while roughly 70% surface more than a month later, usually after a dispute over compensability or medical treatment is already underway. By the time representation appears in the file, the early handling friction that shaped it is months in the past.
What Predictive Modeling Actually Does
This is the gap predictive modeling is built to close, though it is rarely sold that way. The standard pitch is about efficiency, faster routing and lighter workloads, and those are real benefits. The true value of a model lies in its ability to flag, at first notice, which seemingly benign claims are tracking toward a high-cost path while adjusters still have time to intervene.
What a good model adds is the ability to read combinations of data points. A short reporting delay means little by itself, and a soft-tissue injury is the most ordinary thing in the file. The same soft-tissue injury, reported late, in a jurisdiction where attorney involvement runs high, is a different matter altogether, and an experienced adjuster would read it as one. The model applies that reading to every file at first notice, not only the ones a senior adjuster happens to pull.
That experience and intuition matter more every year, because the people who currently supply it are leaving the workforce. Drawing on Bureau of Labor Statistics data, the U.S. Chamber of Commerce projects that about half of insurance professionals aged 55 and older will retire by the early 2030s, opening a gap of more than 400,000 roles.
A senior adjuster carries an internalized sense of how these complex claims resolve, assembled from thousands of outcomes, and that sense is exactly what flags the dangerous file early. A good predictive model captures this intuition and helps improve outcomes across the entire portfolio.
Where This Lands in Practice
None of this pays off without changing the way claims are triaged. Too often, a model is built and wired into an existing system where it produces a credible score, only to have the claim routed to the same queue to sit on the same timeline. As a result, the score becomes another ignored data field on a screen. A predictive score that changes no real-time decision is just an expensive form of reporting.
The programs that produce results do the harder thing, which is to change the work that follows the score. High-propensity claims get a senior adjuster regardless of the dollar reserve, and they get a substantive outreach call inside the first day rather than a form letter in week three. That call is the intervention the WCRI data points to: the claimant's sense of whether the process is fair forms in exactly that window, and it is the strongest lever the carrier has before an attorney is ever involved.
What a bodily injury claim will ultimately cost takes shape over the months after first notice, and the carrier has a hand in shaping it. The carriers that organize around that insight, and use a predictive model to find the files where their hand still matters, are the ones that will keep pulling away from the field. The rest will continue sorting claims by how they look on day one, only to be surprised by what those claims eventually become.
Sources
Insurance Research Council, Attorney Involvement in Auto Injury Claims (2014).
Savych, B., et al. "Unresolved conflict in workers' compensation: The impact of legal representation on workers' compensation benefits." Journal of Risk and Insurance (2026).
Milliman IntelliScript, "How does attorney involvement impact workers' compensation claims?" (2025).
U.S. Chamber of Commerce, The America Works Report: Industry Perspectives.
Workers Compensation Research Institute, research on the drivers of attorney representation.