TL;DR — Stop asking adjusters to simultaneously comprehend thousands of pages and predict financial exposure. Separate the reading from the math by using generative AI to structure the file and geometric machine-learning models to forecast the settlement range.
A catastrophic reserve increase rarely happens on day one. It happens on day four hundred, right before mediation, when defense counsel finally synthesizes the plaintiff's medicals and realizes the venue risk. The initial reserve was set using a severity table from three years ago. That table is now fiction. Social inflation and third-party litigation funding have fundamentally altered the geometry of a claim. The floor for a settlement has shifted upward. The ceiling has disappeared. Yet claims organizations continue to post day-one reserves based on the instinct of an adjuster looking at the first fifty pages of a file. This guarantees reserve volatility. It forces stepping reserves up quarter after quarter, destroying the balance sheet and misallocating defense spend.
The failure is not a lack of skill on the claims desk. It is a capacity limit in the face of an adversarial system designed to overwhelm. A typical litigated claim file quickly balloons to thousands of pages. Pleadings, medical records, endless plaintiff correspondence. The adjuster skims. They look for familiar patterns to justify a number. But the signals of social inflation—a specific plaintiff firm's aggressive tactics, subtle shifts in medical treatment patterns designed to anchor high damages, the unseen presence of litigation funding—are buried deep in the text. Plaintiff attorneys know this. They use document dumps to obscure their strategy. By the time the adjuster or defense counsel recognizes these signals, the claim has escalated. The insurer is playing defense against a narrative the plaintiff has spent months building.
The structural flaw in single-point reserving
The industry response to rising nuclear verdicts has been to mandate higher initial reserves across the board. This is a blunt instrument. It ties up capital unnecessarily on claims that will settle early and often underfunds the actual catastrophic risks. The core failure is treating a reserve as a single, static point rather than a dynamic range of probabilities. When an adjuster writes down a specific number, the organization anchors to it. Negotiation strategies and defense budgets are built around that singular guess. If the guess is wrong, the entire strategy collapses. The plaintiff bar operates on probability and expected value. Insurers operate on static guesses.
To combat social inflation, claims leaders must separate the act of comprehending a file from the act of predicting its cost. Comprehension is a reading problem. Prediction is a math problem. Currently, we ask adjusters to do both simultaneously, using only their experience and limited time. This is where modern forecasting platforms like Canotera intervene. We deploy generative AI to read the case file. It ingests the pleadings, the medicals, the correspondence. It does the reading and structuring of the unstructured data. It isolates the specific drivers of severity—the venue, the injury type, the plaintiff counsel's history. It maps the timeline of medical interventions to detect the artificial inflation of damages.
But the generative AI stops there. It does not predict the outcome. Language models are built to predict the next word, not the financial exposure of a complex casualty claim. Prediction requires a different architecture entirely. We use separate mathematical and geometric machine-learning models to calculate the future. These models are trained on large volumes of resolved cases with known outcomes. They take the structured facts extracted from the current file and map them against the historical reality of how similar claims actually resolved. They measure the distance between the current claim and past claims across dozens of dimensions. This separation of duties prevents the errors inherent in asking a language model to do math. It grounds the forecast in empirical reality.
Calibrating the defense
The result of this architecture is a calibrated output. Instead of a single point guess, the platform provides a settlement range. It identifies the escalation probability. It presents comparable resolved cases. It calculates a reserve delta versus the current posted reserve. Crucially, every number is traceable back to the source documents. The adjuster can see exactly which medical report or pleading drove the forecast. This traceability builds trust. It shifts the adjuster from relying on gut instinct to negotiating from data. When a plaintiff demands an exorbitant sum, the adjuster can counter with a specific range backed by five identical resolved cases from the same jurisdiction.
This changes the operating model entirely. Armed with a calibrated settlement range on day one, claims leaders can allocate defense spend proportionally to the actual risk. A claim with a high escalation probability and a wide settlement range warrants top-tier defense counsel immediately. You do not send a junior associate to fight a plaintiff firm backed by litigation funding. Conversely, a routine claim with a narrow range can be routed to a less expensive path. We stop overpaying to defend claims that should settle quickly, and we stop under-defending claims that carry nuclear risk. The legal budget becomes a strategic tool rather than a sunk cost.
Early detection of escalation is the only reliable defense against social inflation. Plaintiff firms use litigation funding to prolong discovery, build medical specials, and push cases toward trial. They use time as a weapon. If an insurer waits until defense counsel submits a pre-mediation report to understand the true exposure, the battle is already lost. The settlement floor has been established by the plaintiff. The insurer is forced into a reactive posture, paying a premium to avoid the risk of a runaway jury. By identifying the escalation probability on day one, the insurer regains control of the timeline. They can initiate settlement discussions before the plaintiff's sunk costs make early resolution impossible.
By setting realistic reserves on day one, we eliminate the quarter-over-quarter surprises that plague claims departments. We give actuaries accurate data to price the book. We give adjusters the specific comparable cases they need to counter aggressive plaintiff demands. The narrative shifts from relying on historical averages to understanding the specific facts of the case at hand. We dismantle the plaintiff's advantage of time and preparation. The business impact is immediate: lower cycle times, reduced allocated loss adjustment expenses, and a stable balance sheet.
You cannot out-negotiate a reality you refuse to measure.
Related articles.
Why Day-One Reserves Are Systematically Wrong
Setting an initial reserve is an exercise in institutional guesswork. Adjusters are forced to pick a number before the facts materialize, creating a compounding cycle of misallocated defense spend and missed negotiation leverage.
Predicting the Outcome of Disputes: What Claims Leaders Need
Setting initial reserves based on gut instinct and a quick skim of a file is a liability. In an era of social inflation and litigation funding, claims leaders need calibrated forecasts, not wait-and-see guesswork.
From Reactive to Proactive Claims Management
The difference between a reactive claims organization and a proactive one is the trigger for reserving. If you wait for a plaintiff demand or a deposition to understand your exposure, you are already losing the case.
Want to talk to an executive?
Press, partners, investors, candidates — the inbox is monitored. Tell us who you are and we'll route it to the right person within two business days.