How Do You Quantify That Someone Can No Longer Water-Ski With Their Family?

Frank Giaoui spent nearly 30 years as a business and economics consultant — first at Bain, then running his own M&A-focused boutique — before a personal and professional experience roughly 15 years ago sent him back to school, twice. The problem that hooked him intellectually: certain categories of damages, particularly future and non-economic harm like pain and suffering, resist precise quantification in a way material damages (lost wages, medical bills) don’t. His own illustrative example: someone who used to enjoy water skiing with their family every weekend, permanently unable to after an accident — how do you put a number on that loss of enjoyment?

That question led him to a first PhD in comparative law and economics between France and the US (completed at École Supérieure in Paris), followed by a second doctorate — a JSD, essentially a PhD in law — as a visiting scholar and then student at Columbia Law School. The algorithmic research from that second doctorate became the founding basis of Optimalex, incorporated two years before this recording alongside three co-founders (an engineer, a lawyer, and a third partner — one French, one Indian-American), even though the underlying research stretches back over a decade.

In Episode 79 of InsurTechTalk, Frank and I covered how Optimalex clusters litigated cases to predict fair settlement values, why claims frequently end up either under- or over-compensated by default, and the specific loss ratio improvement the company targets with pilot clients.

About Frank Giaoui

Frank Giaoui is Founder and CEO of Optimalex, a legal predictive analytics SaaS company serving insurance carriers, providing claims decision support based on data science, law, and economics. Before Optimalex, he spent nearly 30 years in consulting, first at Bain & Company and then running his own M&A-focused boutique. He holds two doctorates: a PhD in comparative law and economics (France/US) and a JSD from Columbia Law School.

The Core Insight: Cases Aren’t Actually Unique

Optimalex’s foundational legal theory, developed across Frank’s decade of research: despite the conventional legal instinct that “every case is unique,” cases can genuinely be clustered by similar circumstances, venue, and injury type — and once enough data accumulates within those clusters, real, usable patterns emerge. When a new claim arises within a legal situation the model has already learned, the technology compares it against the accumulated pattern of similar past cases to generate a prediction.

The data challenge underneath this: litigated case outcomes are largely public, but the relevant data buried inside them isn’t structured or readily extractable. Optimalex built natural language processing specifically to extract, curate, and structure that data from raw legal filings and outcomes — Frank’s estimate is that only roughly 10% of the total data volume in a given case file is actually useful for prediction, and identifying which 10% matters is itself a core part of the company’s technical work.

What the Product Actually Does

Optimalex sells to insurance carriers, TPAs, and MGAs (with law firms as a plausible but not-yet-pursued customer segment) as a decision-support SaaS tool used by claims managers and adjusters. The workflow: a claims professional can begin using the interface on a new case before it becomes litigated and before a lawyer is even appointed, documenting available information as it accumulates. As more data is entered, the tool’s predictions sharpen — first predicting whether a given case is more likely to settle or proceed to litigation based on circumstances, venue, and injury type, then providing a range of fair settlement values once that determination is made.

Frank’s important clarification on “fair”: it doesn’t mean systematically lower. Optimalex’s core finding is that claims resolution today runs in both directions of error simultaneously — some claims are genuinely under-compensated, while others get pushed unnecessarily far into prolonged settlement or litigation and become significantly over-compensated as a result. The tool’s value is producing consistency, not simply cost reduction — which benefits both the insurer’s loss ratio and, Frank argued, the claimant directly, since under-compensated claimants who might otherwise be pushed into a lengthy, anxious litigation process for uncertain (and often disappointing) results instead receive a fair value faster.

Why SIUs and Claims Organizations Are the Natural First Adopter

I pushed on where in a claims organization this tool creates the most obvious leverage, and the logic tracks directly to cost escalation up the claims pyramid: the longer a claim takes to resolve — moving from initial assessment toward settlement negotiation toward full litigation — the more legal cost, adjuster time, and loss ratio pressure accumulates. A tool that can predict a fair settlement value early, before a case migrates toward expensive litigation, addresses the cost problem at its cheapest possible intervention point.

The TPA Conflict-of-Interest Question

I raised a pointed structural concern directly: TPAs are sometimes compensated in ways that could create an incentive to extend claims duration rather than resolve them quickly. Frank agreed the concern is valid in principle (with an appropriately hedged caveat that he might be wrong about which specific players this applies to), but offered a counter-framing: the most sophisticated TPAs and defense firms increasingly optimize for long-term client relationship value rather than maximizing billable time or duration on any single case — since insurer clients satisfied with efficient claims handling generate more repeat business than any single case’s extended fees would produce. Notably, Optimalex doesn’t currently work with any TPAs — its signed carrier relationships (still under NDA at the time of recording, expected to be announced around ITC Vegas) are directly with mainstream insurance carriers.

The Target: A 10-Point Loss Ratio Improvement

Frank gave a specific, concrete performance target. For a carrier operating around a 70-75% loss ratio — roughly average — Optimalex’s stated goal, achieved over time as the algorithm calibrates to a specific client’s own historical data, is to reduce that loss ratio by roughly 10 percentage points (e.g., 70% down toward 65%, or 75% down toward roughly 68-70%).

The pilot methodology he described: define specific KPIs with each client upfront (loss ratio reduction is common, but he cited a specific client whose primary target was reducing mock trial expenses — a substantial line-item cost per case that Optimalex could largely eliminate). The pilot then runs the algorithm’s predictions against the client’s own historical closed cases to calibrate accuracy, ingests the client’s own proprietary data to sharpen predictions further, and iterates — with a typical pilot running roughly a year before a client commits to full deployment.

Differentiation: Legal Competence, Not Just Predictive Analytics

Frank was direct that Optimalex faces indirect competition from generalist predictive analytics companies, but drew a sharp distinction in approach: Optimalex’s team combines genuine legal training with data science, rather than applying generic predictive modeling to legal outcomes without underlying legal domain expertise. One co-founder holds a PhD in law; another, primarily a data scientist, has spent enough years working directly alongside lawyers to explain legal reasoning in plain terms despite not being formally trained in law himself.

The practical consequence of that combination, in his framing: Optimalex is deliberately cautious about producing predictions that can’t be legally explained or interpreted — the company won’t hand a client a number it can’t justify through underlying legal and economic reasoning, precisely because opaque, unexplainable predictions would undermine the fairness and consistency the product is built to deliver in the first place.

Why Insurers Sit on Data They Can’t Use

Frank’s summary of the actual gap Optimalex fills: prospective and existing clients are typically already extremely data-rich — carriers hold enormous volumes of claims history — but lack the specific technology to identify which portion of that data is actually predictive, or “gold,” as he put it. The combination of legal domain knowledge and data science is what lets Optimalex extract usable signal from data insurers already own but can’t independently mine effectively.

ITC Vegas: A First-Time Attendee, With News Pending

At the time of recording, Optimalex was preparing for its first appearance at ITC Vegas — a meaningful investment for a young company, per Frank — including applying for the ITC pitch presentation (sponsored by State Farm that year) and planning to publicly announce its first carrier partnership around the event. He extended an open invitation for personalized demos and free trial pilots to anyone attending.

Advice: Repeat, and Learn From the Mistake

Asked for closing advice, Frank’s answer was a single word repeated for emphasis: repeat — correct any mistake, and turn failure into the basis for the next iteration. He connected this directly to both his own research process and how Optimalex works with clients operationally: every limitation encountered becomes an asset once addressed through additional data and refined legal analytics, in a continuous, ongoing improvement cycle rather than a fixed, one-time build.

Key Takeaways

• Optimalex’s core legal theory holds that individual cases, despite conventional wisdom, can be meaningfully clustered by circumstance, venue, and injury type — producing genuine predictive patterns rather than requiring case-by-case uniqueness
• Claims resolution today errs in both directions simultaneously — some claims are under-compensated, others over-compensated through unnecessarily prolonged settlement or litigation — meaning the value proposition is consistency and fairness, not simply cost-cutting
• Only roughly 10% of the data within a given legal case file is actually predictive; extracting and structuring that specific subset via NLP is core, non-trivial technical work
• A roughly 10-percentage-point loss ratio improvement is Optimalex’s stated target for carriers implementing the tool over time, calibrated against each client’s own historical claims data during a typical year-long pilot
• Combining genuine legal domain expertise with data science, rather than applying generic predictive modeling alone, is Optimalex’s stated differentiation against indirect predictive analytics competitors
• The company deliberately avoids producing predictions it can’t legally explain or justify to a client, treating explainability as a core design constraint rather than an afterthought
• Carriers are typically data-rich but technology-poor when it comes to identifying which portion of their existing claims data is genuinely predictive — the gap Optimalex is built specifically to close