The Claim Is Decided In The Medical Record And Nobody Has Time To Read It

Thirteen months after a bodily injury claim is denied, the claimant appeals. The reviewing adjuster pulls the original emergency room records and finds soft tissue injuries documented at the exact body part the claimant attributed to the accident. It was on page 547 of a 1,205-page file that the adjuster had four days to review alongside thirty-six other open claims. Nobody was negligent, but when the appeal surfaces that report, it will show exactly how much of that file was read before the claim was closed.

That is the review gap: the distance between the volume of a modern medical record and the realistic capacity to review it. The gap is getting wider every year, and it carries a liability the industry has not priced into its reserves.

What The Data Says

Bodily injury now accounts for roughly 51% of total auto indemnity dollars while representing less than a quarter of claim volume, per CCC's Q2 2025 industry analysis. The average third-party BI payment has climbed about 36% since 2020. Casualty severity is rising around 8% annually - more than double general economic inflation. The value of a claim has become almost entirely attached to the medical record, which makes the quality of the review one of the few real controls a carrier has over loss costs.

That control is slipping.

A single contested injury file routinely runs thousands of pages, pulled from multiple providers, out of order, padded with duplicate documents, imaging reports, and billing ledgers. The shift to electronic health records (EHRs), intended to improve documentation, has instead produced what clinicians call "note bloat." A 2022 analysis of more than 100 million clinical notes found that roughly half of all text in a typical note is copied forward from earlier records. Note length grew about 60% over a single decade, per research published in the National Library of Medicine.

The decisive detail is in the file, but it's surrounded by a thousand near-identical pages that make it almost impossible to find under pressure.

The Workforce Doesn't Grow With The Files

Industry estimates suggest about a quarter of claims adjusters will retire within the next five years. The U.S. Chamber of Commerce projects that half the current insurance workforce will leave within fifteen years, opening more than 400,000 positions the industry cannot fill at its current pace. The reviewers walking out the door are the most experienced ones, the people who knew by instinct where to look. Their replacements are inheriting larger records, higher caseloads, and the same (or shorter) deadlines.

This is not a workflow problem. It's a structural problem that keeps compounding, and while the workforce shifts, the legal teams on the other side are accelerating.

The Reserve Looks Right Until It's a Loss

A missed detail in a 2,000-page file doesn't cause problems that day. The reserve is set, the file keeps moving, and everything goes according to plan. The real consequences show up later, when an oversight becomes a loss.

A pre-existing condition that never got flagged is the causation defense you never got to make. The treatment gap nobody caught is the inconsistency plaintiff's counsel gets to explain first. The specialist's charges, billed at four or five times what any insurer actually paid, become the anchor figure a jury reasons from, which is why Georgia's legislature and Utah's Supreme Court both moved in 2025 to curb these "phantom damages." A reserve built on an incompletely reviewed file is wrong from the day it is entered; everything after that is just the bill arriving.

The downstream numbers reflect it. Nuclear verdicts rose 52% in 2024, reaching a median above $50 million, according to Marathon Strategies' annual analysis. Commercial auto liability sits an estimated $4 to $5 billion under-reserved after fourteen consecutive years of underwriting losses, per AM Best. The medical record is where those numbers start.

The Opposition Has Already Read Your File

Plaintiff attorneys are already using AI to review medical records, surfacing favorable findings, identifying inconsistencies in treatment timelines, building damages arguments before the demand letter is even written. By the time a file lands on your adjuster's desk, attorneys may have already found the four pages in that 3,500-page record that support their theory of damages.

The review gap is a competitive disadvantage, and it is being exploited on the other side of every contested file. Carriers operating on traditional review timelines are working from the same records but at a fraction of the speed and efficiency.

Not All AI In Claims Is The Same - The Distinction Matters

The industry is using AI in claims, but not all tools are built the same way. The difference is where the legal and regulatory risk lives, and the distinction is straightforward: is the AI making the call, or preparing the file so a human can?

When an AI tool makes the call, it becomes discoverable. If opposing counsel can argue that software overrode or displaced an adjuster's judgment, the algorithm's construction, training data, and error rate all become fair game in litigation. That exposure is already materializing in health coverage disputes and legal commentators are tracking its migration into P&C claims.

AI that works on the record rather than the decision is a different proposition. The technology organizes, de-duplicates, indexes chronologically, and surfaces relevant evidence - but it does not score the claim, rate causation, or recommend an outcome. It streamlines the review so the decision stays with the qualified professional reviewing the file.

The NAIC's Model Bulletin on AI use, now adopted in substantially similar form by roughly half the states, makes that accountability clear: a licensed professional owns the claims outcome. The useful test to apply to any AI solution entering the claims space right now is a simple one: does this tool work on the record, or does it work on the decision?

How It Works In Practice

SiftMed works on the record. The platform takes complex, disorganized medical files and turns them into a structured, searchable record - organized, de-duplicated, and indexed, with relevant evidence surfaced for the reviewer to weigh - so they're working with the full context from the first page. SiftMed doesn't score the file or suggest a decision; it makes the record navigable. Across our customer base, users cut medical record review time by 50 to 70%.

The time saving is a byproduct; the real return is what recovered hours buy.

An adjuster reviewing a 2,000-page file might spend the first two to three hours on document navigation - sorting by provider, finding the date range, identifying what they are even looking at - before forming a single clinical judgment. With SiftMed, they open a structured, de-duplicated record and spend those same hours on actual review: reading the clinical notes, assessing causation, finding the inconsistencies that move the scale. The AI never touched the decision.

That distinction matters for a legally cautious audience because it is auditable. The record log will show that a qualified human had an organized file and the critical details in front of them before the reserve was set. That is not just an efficiency claim. It's a defensibility claim, one that means something different when a file goes to litigation and opposing counsel starts asking what the adjuster actually reviewed.

Before You Move The Next File

When your next decision is analyzed, what will the record review show? If a file goes to litigation and opposing counsel asks how much of the medical record was reviewed before the reserve was set - what is the answer, and will it hold up?

Those are not hypothetical questions. They are the ones worth asking before the next file closes.

The carriers that come through this decade with their loss costs intact will not be the ones with the most adjusters or the thickest checklists. They will be the ones who put capable AI on the record itself while keeping a licensed human in the chair where the medicine, the law, and the regulators all expect one. Reading the file faster was never the point, what matters is what gets discovered inside it.

Find out what's in your next file before opposing counsel does, see SiftMed on your own files. Request a demo.