Allstate Patents a Catastrophe Claims Engine That Pre-Stages Adjusters

Most catastrophe patents stop at the loss number. US Patent 12,700,045, granted to Allstate Insurance Company on August 4, 2026, does not: it forecasts claim volume from real-time sensor and weather data, then continuously reallocates adjusters against capacity thresholds and an overflow queue (USPTO Official Gazette, week 31, 2026). That reframes loss adjustment expense as a modeled, cat-driven quantity rather than a flat percentage load.

Key Takeaways

• The fourth grant in a family filed since March 21, 2011. What the 2026 version adds is a closed loop: continuous workload monitoring that automatically pulls claims back out of the overflow queue the moment an adjuster frees up.
• Live claim count and queue dwell time are the two inputs the CAS's claim-count-weighted alternative to paid-to-paid ULAE estimation has always been starved of for catastrophe events. This system generates both at the moment of the event.
• $1,722 million of catastrophe losses in Allstate's second quarter of 2026, against a property-liability combined ratio of 86.6 helped by 2.4 points of favorable catastrophe experience.
• Only the labor half of demand surge is addressed. The patent does nothing for lumber prices or contractor day rates, so it is not evidence that demand-surge severity trend should moderate.

Patent Details

Patent Number	U.S. 12,700,045 B2
Filed	December 11, 2023 (priority to March 21, 2011)
Granted	August 4, 2026 (Official Gazette week 31)
Assignee	Allstate Insurance Company
Sensor inputs named in Claim 1	Wind speed, rainfall, fire speed, water speed, water depth, earthquake magnitude
Family	Fourth grant in the "Claims Adjuster Allocation" lineage, after US 9,947,050 and US 11,842,405 B1

A Fifteen-Year-Old Family Gets a Closed Loop

"Claims Adjuster Allocation" is not a new idea at Allstate. The family has been re-filed and re-granted three times since 2011: a first patent in the lineage, then a continuation, then US Patent 11,842,405 B1, filed August 14, 2020 and granted December 12, 2023, and now US 12,700,045 B2.

Every version shares the same core mechanism. A central adjuster control unit monitors sensor data to identify a region where a catastrophic event may occur, estimates the resulting claim count, and assigns claims adjusters against a threshold capacity per person or workgroup.

What is new in the August 2026 grant is the closed loop at the back end. The independent claim requires "continuously monitoring, by the processor, workloads of the plurality of claims adjusters to detect when a claims adjuster has capacity below the threshold number," and, once that happens, "automatically re-assigning, by the processor, one or more of the insurance-related claims from the capacity entity to the claims adjuster up to the threshold number" (US Patent 12,700,045 B2, claim 1).

In the 2023 parent, the overflow bucket that the claim language calls a "capacity entity" was closer to a static holding pen: claims waited there until someone checked capacity and pulled them out manually. The 2026 claim makes that pull automatic and continuous. That is the difference between a surge plan a supervisor runs once a day and a staffing algorithm that rebalances itself in real time.

That specificity is also why the family keeps clearing examination. Claims on using data to allocate resources sound like the abstract business methods invalidated under Section 101 since Alice, but this one recites the sensor inputs by name, the threshold mechanism, the capacity entity as a memory structure and the monitoring loop that triggers reassignment. It matches the pattern the site found in the USPTO's 2026 examiner guidance reset: narrow claims tied to a specific technical sequence survive scrutiny that broad "use AI to route work" claims do not.

LAE as a Forecast, Not a Load

Most reserving actuaries still carry loss adjustment expense as a flat load, a percentage applied to indicated losses, calibrated from a paid-to-paid ratio of calendar-year claims department expense to calendar-year paid losses. The Casualty Actuarial Society's own literature flags the limit: paid-to-paid does not adjust well when claim complexity or claim count shifts sharply from one period to the next, which is exactly what a catastrophe quarter does (CAS).

The proposed alternative ties ULAE to a weighted count of claims still to be settled, and it has always been data-starved for cat events because nobody had a live count of claims still coming. US 12,700,045 solves exactly that, even though nothing in it mentions reserving. The system estimates potential claims in a region before the event fully develops and tracks in real time how many are assigned against how many sit in the capacity entity.

The difference is concrete. Under paid-to-paid an actuary applies, say, a 12% ULAE load to indicated cat losses because that is the trailing three-year average, regardless of whether the event produced 4,000 claims or 40,000. Under a claim-count-weighted approach the same actuary multiplies a per-claim handling cost by expected claim count, then adjusts upward for claims routed through the overflow queue, because a claim that waits is by construction taking longer and costing more to close.

Cycle time moves with it. Milliman's work on demand surge found hurricane claims already develop more slowly on incurred loss and allocated expense than standard homeowners claims, driven partly by the strain of bringing in out-of-area adjusters under the ad hoc process this patent replaces (Milliman). A system built to keep every adjuster near threshold should pull that development curve forward.

An actuary applying an unadjusted, pre-2026 LDF selection to a cat event handled this way risks understating near-term paid loss, and misjudging when a quarter's cat losses have finished emerging. Allstate's own quarter shows the scale the load sits against: $1,722 million of catastrophe losses, down from $1,990 million, with the property-liability combined ratio improving 4.5 points to 86.6 (Form 10-Q, Q2 2026).

The Half of Demand Surge It Does Not Touch

Demand surge has two bottlenecks and this patent attacks one. A claims engine that forecasts volume and reallocates adjusters reduces the odds an insurer is caught short on adjuster headcount at landfall. It does nothing for lumber prices, contractor day rates, or roofing material shortages, which are set by the regional construction market.

A reserving actuary treating the patent's existence as evidence that demand-surge severity trend should moderate would be conflating the two halves. Only the claims-handling bottleneck is patented here; the materials-cost bottleneck responds to entirely different levers, and in a year the site's tracking already put past $35 billion in insured US losses by August from wind and hail events alone, that second bottleneck is the one still setting severity.

There is a subtler exposure in the convergence. The sensor and weather inputs named in claim 1 are the same live exposure data the site has tracked moving into real-time probable maximum loss monitoring on the pricing and capital side. One feed now drives both a rate filing's cat load and a claims-operations staffing decision.

Reconciling those against a single underlying feed is an improvement over two independently built estimates that quietly disagree. It also means a single modeling error propagates into pricing and into LAE at once, with no second estimate left to contradict it.