- Event
- ITC Vegas 2026 (InsureTech Connect), expo floor
- Venue
- Mandalay Bay, Las Vegas, Nevada
- Date
- Thursday, October 1, 2026
- Spoke with
- Mubbin Rabbani, Chief Product Officer, CLARA Analytics
Key takeaways
- CLARA brings AI agents and analytics to claims, mainly workers’ compensation and commercial casualty, plus personal auto.
- The goal is to “get to risk faster”: spot the claims likely to turn into litigation, lost time or long-term medical costs right after first notice of loss.
- Instead of reviewing files on a fixed cycle, adjusters are pointed to the claim that needs them today.
- Rabbani said clients typically see a measured 2 to 5 percent impact on losses within about a year, checked with their own actuaries.
- Audit logs, runbooks, bias and accuracy controls, and data governance come first, before any user goes live.
What does CLARA Analytics do?
CLARA Analytics, based in Sunnyvale, California, and led by CEO Heather H. Wilson, builds AI for insurance claims. Rabbani said the focus is workers’ compensation and commercial lines such as auto liability and general liability, in both primary and excess layers, as well as personal auto.
The timing was good. On September 30, 2026, the day before we spoke, CLARA announced an agentic layer for its CLARAty.ai platform, which it describes as an AI workforce that supports claims from first notice of loss to resolution while adjusters stay in control of the decisions.
How does AI change an adjuster’s day?
Rabbani started with a simple problem. An adjuster may have hundreds of claims open at once. New information keeps arriving, but in a traditional model, he said, an adjuster could “cycle out weeks before they touch that claim again.”
CLARA pulls in data from a carrier’s claim, legal, medical and document systems, including free text. It turns all of it into one picture of the risk ahead. Right after first notice of loss, it watches for warning signs. These include the chance of litigation or a difficult jurisdiction on a bodily injury claim. On a workers’ comp claim, it looks for the chance of lost time and long-term medical costs.
The aim, in his words, is to get “the right adjuster to the right claim at the right time” so they can step in before a claim drifts. He said clients usually measure the result with their own actuaries and analysts, and see a 2 to 5 percent impact on losses, typically within a year.
How do you trust what the agents are doing?
This was the question that mattered most to me, and Rabbani answered it directly. Before CLARA switches anything on, he said, the team works with each client to make sure its data is “ready for AI.” They check where the data comes from, how good it is, what format it is in, and how the agents will treat it.
He described full audit logs, controls and runbooks for the agents. A carrier can see what the agents and models are doing and what they find. That includes bias controls and accuracy controls. The key point, he said, is having data governance, AI governance, model audits and data security in place “before we even deploy a single user onto our platform.”
What happens to the claims data?
Rabbani said CLARA runs the data through a pipeline that removes personal identifiers before it joins a larger shared model. That pooled, anonymized data does two jobs: it helps train and tune CLARA’s models, and it gives every client insights none of them could build alone.
He gave two examples. One product tracks how plaintiff attorneys and law firms perform on certain types of claims and in certain venues. He said no single industry data set offers this today. Another, in workers’ comp, looks at how treatment progresses with specific doctors, to help find the path that gets injured workers back to work.
Why it matters for insurers
Most claims leaders do not need to be sold on AI any more. What they need is proof that it changes loss costs, and confidence that it can be governed. Rabbani’s answer put both on the table: outcomes measured by the client’s own actuaries, and controls, logs and runbooks set up before anyone uses the system.
ZERO LEGACY, from Zero Legacy Press, makes the same point about agents: the model alone is not the product. As the book records, “the model is the brain, the harness is the body.” The work that decides whether AI can be trusted in claims is the harness around it. Read about the book
Who we spoke with
- Mubbin Rabbani, Chief Product Officer, CLARA Analytics