Insurance claims leakage: how uninvestigated claims drain profitability

What is claims leakage

Claims leakage is the difference between what an insurer actually pays on a claim and what it should have paid given proper investigation, accurate valuation, and correct coverage application. It is money that leaves the organization unnecessarily - not through catastrophic one-time events, but through a steady accumulation of overpayments, missed fraud, overlooked subrogation opportunities, and avoidable settlement inflation across thousands of claims.

McKinsey estimates that claims leakage represents 5-10% of total claims expenditure across the P&C industry. For an industry that pays out over $300 billion in claims annually in the US alone, that translates to $30 billion or more in avoidable losses every year. These are not catastrophic losses from a single event - they are systemic losses embedded in the claims handling process itself.

What makes claims leakage particularly insidious is its invisibility. Unlike fraud - which is at least theoretically detectable - leakage often occurs within the normal range of claims handling outcomes. An adjuster who settles a claim for $15,000 when proper investigation would have revealed a $9,000 valuation is not committing fraud. They are making a reasonable-seeming decision with incomplete information. But multiplied across thousands of claims, these reasonable-seeming decisions cost billions.

Root causes of claims leakage

Claims leakage has multiple root causes, but they share a common theme: insufficient investigation depth. When adjusters and investigators do not have the time, data, or tools to fully evaluate a claim before making a payment decision, overpayment is the predictable outcome.

1. Uninvestigated fraud

The most direct source of leakage is fraud that goes undetected because the claim was never investigated. As we detailed in our analysis of why 75% of flagged claims are never investigated, SIU teams lack the capacity to investigate most of the claims that their scoring models flag as suspicious. The flagged claims that never reach an investigator represent pure leakage - the insurer knows these claims warrant scrutiny but pays them anyway because there is no one available to do the work.

The financial impact is significant. If 10% of claims are flagged as suspicious and only 25% of those receive full investigation, the insurer is paying 7.5% of its flagged claims without scrutiny. If even half of those uninvestigated flagged claims contain some degree of fraud or overpayment, the leakage from this single source can represent 3-4% of total claims expenditure.

2. Untrained or overloaded adjusters

Claims adjusters are the first line of defense against leakage, but they operate under significant pressure. Average caseloads at many carriers have increased steadily over the past decade as insurers pursue operational efficiency through headcount reduction. An adjuster managing 150-200 active claims simultaneously cannot conduct thorough investigation on every file.

The result is pattern-based decision-making. Adjusters develop heuristics - rules of thumb for valuing claims based on similar past cases. These heuristics are generally reasonable but they systematically favor overpayment because the professional and regulatory consequences of underpayment (bad faith claims, regulatory action) are more severe than the consequences of modest overpayment (which is largely invisible).

3. Missed subrogation opportunities

Subrogation - recovering claim payments from a responsible third party - is a significant revenue source for P&C insurers. But subrogation requires identifying the responsible party, establishing liability, and pursuing recovery - all of which require investigation time that overburdened adjusters often cannot dedicate.

Industry estimates suggest that 20-30% of subrogation-eligible claims are never pursued, either because the subrogation opportunity was not identified or because the adjuster prioritized new claims over recovery efforts on closed ones. At an average recovery of $3,000-$5,000 per subrogation action, the aggregate missed recovery across a large book of business is substantial.

4. Settlement overpayment

Settlement overpayment occurs when a claim is settled for more than its actual value due to incomplete information, negotiation pressure, or litigation avoidance. This is distinct from fraud - the claimant may not be acting deceptively. The overpayment results from the insurer's failure to fully investigate and accurately value the claim before entering settlement negotiations.

Common drivers include: accepting medical bills at face value without auditing for duplicate charges or unnecessary procedures, settling bodily injury claims without independent medical evaluation, and settling litigated claims at inflated values to avoid defense costs. Each individual overpayment may be modest - $2,000-$5,000 on a $20,000 claim - but the cumulative impact is measured in billions.

The investigation gap

The central driver of claims leakage is the investigation gap - the chasm between the number of claims that warrant investigation and the number that actually receive it. This gap exists because the economics of manual investigation make comprehensive coverage impossible.

The math is straightforward. A typical SIU investigator handles 200+ active cases. Each case requires an average of 14 or more days of investigation time. At that rate, an investigator can complete roughly 15-20 full investigations per month. If an insurer processes 50,000 claims per month and 10% are flagged, that is 5,000 claims requiring investigation. To investigate all of them would require 250-330 investigators - a staffing level that no mid-size carrier can justify economically.

The result is triage. SIU teams focus on the highest-value, highest-confidence fraud referrals and let the rest go uninvestigated. This is a rational response to a resource constraint, but it creates a predictable and exploitable pattern: sophisticated fraudsters learn the SIU's capacity threshold and structure their fraud to stay just below it. Claims that are suspicious but not obviously fraudulent - the gray zone - are almost never investigated.

“The claims leakage problem is not a detection problem. Scoring models flag suspicious claims with reasonable accuracy. It is an investigation problem. Flagging without investigating is the same as not flagging at all - the money still leaves.”

- Hesper AI Research, Q1 2026

How AI reduces leakage

AI investigation agents address claims leakage at its root cause: the investigation gap. By automating the labor-intensive steps of claims investigation - document collection, cross-referencing, OSINT, network analysis, and report synthesis - AI agents compress the investigation timeline from 14+ days to 2-4 hours per case. This is not a marginal improvement. It is a structural change that makes investigating every flagged claim economically feasible. See how the Hesper AI investigation agent works.

The impact on leakage comes from four mechanisms:

1. 100% investigation coverage - every claim flagged by scoring models receives a full investigation, eliminating the gray zone where most leakage occurs
2. Faster investigation cycle - claims are investigated within hours of flagging rather than sitting in a queue for weeks, reducing the time window for evidence destruction or witness coaching
3. Consistent investigation depth - AI agents follow the same comprehensive playbook on every claim, eliminating the variability in investigation quality that comes from different investigators with different experience levels and caseload pressures
4. Network-level analysis - AI agents cross-reference every claim against the insurer's full claims history, surfacing patterns (provider networks, claimant recurrence, geographic clustering) that are invisible to case-by-case manual investigation

The first mechanism is the most important. Most claims leakage does not come from investigated claims where the investigator missed something. It comes from claims that were never investigated at all. By making investigation coverage comprehensive rather than selective, AI agents eliminate the single largest source of leakage. Learn how the investigation pipeline works.

ROI of investigation automation

The ROI of AI-powered investigation automation can be modeled directly from the claims leakage reduction it enables. The calculation involves three variables: the number of additional claims investigated (previously uninvestigated flagged claims), the expected fraud or overpayment rate among those claims, and the average recovery per investigated claim.

Here is a realistic model for a mid-size P&C carrier processing 100,000 claims annually:

In this model, the net impact is approximately $26.75 million annually - $22 million in incremental fraud and overpayment recovery plus $4.75 million in SIU cost reduction. The actual figures will vary by carrier size, line of business, and current investigation coverage, but the structural economics are consistent: AI investigation pays for itself many times over because the cost per investigation ($150) is a fraction of the average recovery per confirmed case ($8,000).

It is worth emphasizing that AI investigation does not replace SIU investigators. It augments them. AI agents handle the data-gathering, cross-referencing, and initial analysis that consume most of the investigator's time. Investigators focus on high-complexity cases, witness interviews, and decisions that require human judgment. The result is a more effective SIU team, not a smaller one. Compare this approach to traditional SIU tools.

For carriers evaluating the ROI of investigation automation, the critical question is not whether AI investigation reduces leakage - the economics are clear. The question is how quickly the system can be deployed and integrated into existing claims workflows. Explore use cases by line of business.

Frequently asked questions