AllegoryOS Founder & CEO Onur Gungor Joins Profiles in Risk

I joined Tony Cañas on Episode 862 of Profiles in Risk to discuss what we are building with AllegoryOS, why insurance software needs to work differently, and how my actuarial background shaped our approach to AI.

Watch the full conversation on YouTube.

AI through an actuarial lens

Actuarial work is not only about building models. It requires understanding why data produces a particular signal and how changes in behaviour, demand, supply, and risk affect an insurance portfolio.

That perspective influenced how I approached AI. Insurance pricing, reserving, capital management, and financial planning are all probabilistic. Modern AI models are also probabilistic systems. The challenge is to understand their errors, reduce them, and make their outputs reliable enough for real insurance operations.

I never viewed AI as merely a chatbot. I wanted to understand how it could safely support mission-critical work across pricing, underwriting, claims, policy administration, and software development.

Insurance software starts with data architecture

An insurance company’s most valuable operational asset is its data. Yet critical information is often distributed across spreadsheets, disconnected databases, reports, and systems that do not share a common language.

Before AI agents can perform useful work, the underlying data needs clear definitions and relationships. A policy number, coverage limit, claim status, underwriting rule, interface field, and database table are all data points. A proper data dictionary gives those points consistent meaning across the organization.

Once those foundations are in place, AI becomes a practical operating tool. Teams can explain an operational problem in their own language, and the system can translate that knowledge into requirements and working software much faster than a traditional implementation cycle.

The synthetic insurance universe

We also discussed Allegory’s synthetic insurance universe and the actuarial ideas behind it.

The approach applies simulation and probabilistic modelling to insurance concepts such as deductibles, limits, claims, properties, vehicles, and policyholders. Each concept can be examined across products and operating contexts, creating a structured representation of how insurance works.

That structure supports smaller, insurance-focused models. Instead of asking a general-purpose model to infer every relationship from an enormous collection of unrelated information, we can give the model a defined insurance vocabulary and a more disciplined operating context.

Giving consumers a better way to understand insurance

The conversation also covered the Allegory digital wallet. It allows people to store their policies and ask questions about their coverage, premiums, and claims process.

People do not want another complicated insurance portal. They want help within the channels they already use. In many markets, messaging platforms such as WhatsApp contain more practical information about a customer’s insurance needs than the insurer’s own systems because customers and brokers already use them to exchange documents and resolve problems.

Insurance technology should meet people where they are. It should make policies easier to understand and help insurers support customers when that support matters.

Six months later

We recorded this conversation about six months before it was published. Listening back was also an opportunity to reflect on how much AllegoryOS and our Premium Under Management have grown since then.

The central ideas remain the same: insurance needs better data foundations, AI needs disciplined probabilistic thinking, and software should adapt to insurance operations instead of forcing insurance operations to adapt to software.

Thank you, Tony, for having me on Profiles in Risk.

Watch Episode 862 on YouTube.