Calibrated Reserves: A Better Day-One Number

TL;DR — Moving from a single-point estimate to a calibrated settlement range allows claims teams to allocate defense spend properly and avoid late-stage surprises driven by social inflation.

The day-one reserve is the original sin of a casualty claim. An adjuster gets a new file containing a pleading, initial medical records, and a police report. They skim the documents, pattern-match against their recent experience, and enter a number into the system. That number becomes the anchor. It dictates how aggressively the defense is funded, which tier of counsel is assigned, and whether the file gets escalated to a senior examiner. Most of the time, that number is wrong.

We pretend claims have a singular true value waiting to be uncovered. They do not. A claim is a distribution of possible outcomes. When we force a claims professional to pick a single point estimate on day one, we are forcing them to hide the volatility inherent in the file. If a bodily injury claim looks routine, the reserve is set to a familiar baseline. The system completely ignores the statistical probability that a third-party litigation funder is involved, the venue is hostile, or the plaintiff's counsel has a history of pushing these specific injuries to trial. The point estimate creates a false sense of security that permeates the entire lifecycle of the claim.

The Problem with Point Estimates

The consequences of this false security compound over the life of the file. Under-reserved files get under-defended. A low day-one number signals to the defense team that this is a routine matter. Minimal resources are allocated. Discovery is treated as a formality. Opposing counsel operates with an entirely different playbook. They are building a case for a nuclear verdict while the defense is managing a spreadsheet. Months or years pass. By the time the plaintiff's true strategy emerges, the defense is caught flat-footed.

The reserve then has to be stepped up. Late-stage reserve increases create intense friction between the claims floor and the actuarial department. They cause severe reserve volatility, erode trust with the executive team, and force the carrier to negotiate from a corner. You are no longer managing the claim; the claim is managing you.

You cannot fix this by simply telling adjusters to be more conservative. Broadly inflating day-one reserves ties up massive amounts of capital. It restricts the carrier's ability to write new business and distorts the broader financial picture. The solution requires a structural change in how we define a reserve. We have to move away from the single point estimate and adopt a calibrated loss reserve. A calibrated reserve does not pretend to know the exact final penny. It provides a settlement range. It acknowledges the inherent uncertainty in litigation and quantifies it based on historical precedent.

Building a Calibrated Range

Creating a calibrated range requires separating the act of reading from the act of predicting. The traditional workflow conflates the two. An adjuster tries to synthesize thousands of pages of text while simultaneously guessing the financial outcome. We approach this differently at Canotera. Generative AI is deployed strictly to read. It processes the unstructured mess of the case file. It pulls the specific facts from pleadings, medical records, and dense legal correspondence. It does the tedious work of structuring the reality of the claim. It does not predict the outcome.

The prediction is handled by separate mathematical models. These models do not rely on gut instinct or the limited memory of a single human being. They are trained on large volumes of resolved cases with known outcomes. By comparing the structured facts of the new claim against this historical baseline, the model generates a calibrated settlement range. It calculates an escalation probability. It surfaces comparable resolved cases. It identifies the reserve delta between the model's output and the current manual reserve. This separates the subjective analysis of the adjuster from the objective reality of the market.

Crucially, this output is traceable. A mathematical range is useless if the claims team cannot understand why it exists. The drivers behind the numbers must link directly back to the source documents. If the model flags a high escalation probability, the adjuster needs to see the specific medical finding, the history of the plaintiff's firm, or the venue characteristic driving that risk. This transparency is mandatory. Claims professionals are rightly skeptical of opaque systems. You build trust on the claims floor by providing a tool that explains its reasoning and points directly to the evidence in the file.

Operating on Reality

When you equip a claims team with a calibrated range, the operating model shifts. The day-one number ceases to be a defensive guess. It becomes a strategic baseline. If the upper end of the calibrated range is high, the claims manager knows immediately that this file requires aggressive defense funding. They allocate the right resources early. They assign specialized counsel rather than a high-volume defense firm. They escalate the file to a senior handler before the plaintiff's attorney has a chance to build momentum. The defense strategy is calibrated to the actual risk, not the median expectation.

This changes the dynamic of the negotiation. An adjuster armed with a point estimate is arguing against the plaintiff's point estimate. It devolves into a battle of competing opinions. An adjuster armed with a calibrated range, comparable resolved cases, and traceable data drivers is negotiating from objective reality. They can see the historical baseline for similar injuries in similar venues. They know exactly why the claim is worth what it is worth. They can detect the early signs of a nuclear verdict trajectory and settle the claim before the plaintiff's demands become untethered from reality.

The goal is not to replace the claims professional's judgment, but to give that judgment a foundation built on historical fact rather than isolated experience. We must stop pretending that complex litigation resolves to a single, predictable number on day one. A point estimate is a wish. A calibrated range is a strategy.