Key takeaways
- At 100 to 150 active claims, an adjuster has minutes per file per day, and subrogation review is the first task deferred.
- Over 80% of AI-identified subrogation opportunities come from narrative text, which is exactly what a busy adjuster skims.
- A subrogation question at file closure arrives too late; the signal is usually in the file within the first week.
- The fix is structural: screen every file automatically, route scored leads to the subrogation unit, and sweep closed files on a schedule.
The arithmetic of a full desk
Adjusters carry 100 to 150 active claims at once (Carrier Claims Workload Study). Take the midpoint of 125 files and an eight-hour day. Even if an adjuster spent the entire day on file work, with no calls, meetings, inspections or training, each open claim would get under four minutes. Real days have all of those things, so most files get no attention on most days and a handful get all of it.
On that desk, adjusters triage by urgency: coverage questions, injury exposure, customer complaints, rental days, regulatory response times. Subrogation is rarely urgent in week one. The limitation period is years away, the insured is not calling about it, and nobody's scorecard measures it in real time. So it waits, and on a desk that never empties, waiting often means never.
None of this is a failure of individual adjusters. It is the predictable result of asking one person to resolve a claim and also audit it for recovery while carrying 124 others.
Where the signals hide
Over 80% of AI-identified subrogation opportunities are triggered by unstructured narrative text, not structured claim fields. The structured fields that drive referral rules, such as loss cause codes, a third-party checkbox or an at-fault indicator, are filled in quickly at FNOL and rarely corrected. The facts that create a recovery arrive later and land in the notes.
A water loss coded as sudden and accidental discharge gets a diary entry on day six: 'Insd says plumber replaced the dishwasher line in April.' An auto file coded as a single-vehicle loss gets a recorded-statement summary: 'IV swerved to avoid box truck that crossed center line, truck did not stop.' A commercial fire coded as electrical gets an origin and cause report that places the origin at a tenant's space heater. Each sentence is a recovery lead. None of them changes a structured field, so none of them trips a rule.
The same is true of attachments. A police report narrative cites the other driver for following too closely. An engineer's report attributes a collapsed rooftop unit to missing fasteners from a recent HVAC replacement. A repair estimate lists a component that is under recall. An adjuster on a 125-file desk opens these documents to confirm damages and coverage, which is a different reading from the one a recovery professional would give them.
An adjuster reading for coverage facts reads past these lines, not out of carelessness, but because the question in their head is whether the loss is covered, not who else caused it.
Why the closing checklist is too late
Many carriers rely on a subrogation question at file closure: is there subrogation potential, yes or no? At closing, the adjuster is trying to clear the file, and the honest answer to whether every note was read for liability signals is usually no. Closure can also come months after the evidence was available. By then the failed supply line has been discarded, the intersection video has been recorded over, and the adverse carrier has closed its own file.
Timing drives cycle time as well. Recovery takes about 14 months in manual workflows versus 4.5 months with automated demand packaging and early evidence preservation, and much of that gap opens before a subrogation professional ever sees the file.
Handoffs compound it. Files move between adjusters during catastrophe surges, reassignments and departures, and the new adjuster inherits a diary written by someone else in someone else's shorthand. A lead noted once on day two is rarely restated.
What a blind spot costs
Between 15% and 30% of viable third-party subrogation opportunities are never identified or pursued across US P&C lines (NASP & Verisk / ISO), and US carriers lose over $15 billion a year to missed or abandoned subrogation (P&C Actuarial Review). At the file level, the average missed recovery is $8,400 on a personal auto collision, $28,500 on a commercial property and water loss, and $42,000 on a homeowners fire or product defect claim.
At the carrier level the arithmetic is direct. A $150 million DWP carrier closing about 30,000 claims over 24 months, with a 2.2% yield of unpursued claims, has about 660 files averaging $12,500, or $8.25 million sitting in closed files. Recovering 60% of that before statutes run returns $4.95 million to surplus.
Spread across a staff of 60 adjusters, those 660 files are 11 missed leads per adjuster over two years, or roughly one every two months. That is not negligence. It is one sentence in one diary note, missed on a desk of 125 open claims.
Fix the process, not the people
Adding checklist questions to a full desk does not work, and neither does asking adjusters to think harder about subrogation. What works is separating detection from adjusting. Every file, open or closed, should be read for liability signals by something other than the person trying to close it, and the leads should go to someone whose job is recovery.
In practice that means three changes. Screen open claims continuously as notes, reports and estimates arrive, and route scored leads to the subrogation unit in the first days of the claim. Give subrogation professionals a ranked queue that combines recoverable dollars, win probability and days until the limitation or notice deadline, so they work the files most likely to pay before they expire. And sweep closed files on a schedule, because some facts only become clear after the claim resolves.
Measure it differently too. Useful metrics are the median days from FNOL to subrogation referral, the share of claims screened before closure, the referral rate by unit, and the count of opportunities lost to an expired deadline. A referral rate that varies widely between units with similar books is a caseload signal, not a talent signal.
A simple test shows whether the process works. Pull 50 files closed without recovery last quarter and have a subrogation professional read every note and attachment. Count the files with a viable lead that was never referred, and record the day of the claim on which the signal first appeared. When the signal arrived in week one and the referral never came, the problem is routing, not judgment, and more training will not fix it.
Where Ishqkala fits
Ishqkala's Real-Time FNOL Triggers read each new diary note, recorded statement summary, police report and estimate on open claims and send liability signals to the subrogation unit with the sentence that triggered them. The Closed-Claim Sweeper covers the back book, scanning up to 50,000 closed claims in 48 hours and producing an itemized Recovery Opportunity Manifest. Adjusters keep adjusting, and subrogation professionals get a queue they can act on.
Across Ishqkala customers, that approach has identified more than $45 million in missed recoveries in closed files and prevented more than 12,000 SOL expirations.