Cogitator POV: Humans in the Lead, Friction on the Run

A conversation with Arvind Kaushal, CEO & Co-Founder of Cogitate, on the future of AI in insurance and where Cogitate is headed.

Ask Arvind Kaushal where insurance technology is headed, and he won’t start with a product. He’ll start with the lifecycle, the sequence every policy and every claim follows, and then explain why that sequence is exactly where artificial intelligence is about to do its most important work. In this conversation, the CEO and co-founder of Cogitate lays out a clear-eyed view of agentic AI, the orchestration layer he believes will define the next era of core systems.

Let’s start with the big picture. When someone walks up to you at a conference and asks “Who is Cogitate?” how do you respond?

That is a great question. We start from a conviction: insurance runs on judgment, and we built CAIRA, our AI engine to amplify it, never to replace it. The proof behind that conviction is real: Cogitate is a digital-ready ecosystem offering policy, claims, billing, and distribution, all integrated on a modern SaaS platform that has been running for years. That is the foundation, with billions of dollars of transactions that have passed through it. What we’re doing now is taking that whole lifecycle and putting one intelligence, CAIRA, across that entire lifecycle to take the friction out and keep the expert in the lead. So the question of “who is Cogitate” is rapidly evolving. I want us to be known for what we’re becoming, not only what we’ve already built.

You focus on the word “lifecycle.” Why is that the right frame for understanding AI in insurance?

Let’s look at the policy lifecycle. If I need to do something with a policy, the first thing I need to know is your risk attributes, as that’s how any policy starts. Whether it’s auto, property, or cyber, it begins with a submission, and a submission is really risk data. Once we have that risk data, the question becomes: how do we reduce the friction across the rest of that journey – indication, quote, binder, endorsement, cancellation, reinstatement, integration with payments? Every policy follows pretty much the same lifecycle. The claim starts the moment someone picks up a phone or submits the FNOL via email. Both are entry points to two different lifecycles, and our job is to take the friction out of those workflows wherever it makes the most sense. And that is where CAIRA comes in: one intelligence, showing up at every one of those moments across these lifecycles.

Digitalization has been the story in insurance for years. How is what you’re describing different from that?

Digitalization was one layer. It elevated user experience and improved efficiency. That base isn’t changing. Even years down the line, if someone wants home insurance, it’s always going to start with the same questions: where is your home, what type is it, and in which zip code do you reside? Those things will always be needed. What changes is how we process that information and how data moves through the lifecycle. That movement will be managed by both AI and humans. Gradually, AI will play a bigger role in managing those lifecycles. What we’re doing is augmenting intelligence via CAIRA into that lifecycle to reduce friction expressed as AI agents at each moment that the work happens.

So when you talk about AI agents, let’s expand on how AI differs from the logic insurers have always had?

Think of it this way, there are three pieces are working together. There’s a core AI-based intelligence that solves one specific problem. Then there’s software attached to that agent to make it meaningful for the business. And then there’s a human. Those three working in conjunction create business value. An AI agent on its own doesn’t produce value. Take CAIRA’s inspection agent. By itself, it does exactly one task, and it doesn’t matter whether it’s personal lines or commercial lines. To make the inspection agent work for an Underwriter, there’s a workflow around it that has to be managed, a dashboard for transparency, and a human observing and making final decisions. That combination amplifies the expert, the Underwriter, for an extremely productive experience.

Can you give us a concrete example of one of these agents in action?

CAIRA’s submission agent is a good example. A submission might be fifteen pages, or for a large account it could be five hundred pages, with multiple types of documents like risk data, loss runs, financial information, and so on. The agent reads through all of it, figures out which attributes matter, and intelligently infers from various documents. Our agent goes to real depth here, checking on the order of 140-plus attributes, pulling out the data that matters, synthesizing it and prefilling it into the underwriter’s workbench, ultimately with a recommendation to quote. In other words, the agent clears the toil by reading the pages, extracting and inferring the risk data, filling the workbench with a confidence score and a recommendation to quote. It simply augments or “amplifies” the Underwriter. The decision never leaves the human. In our CAIRA framework, we’ve brought a lot of transparency to the embedded AI. We show our customers how many attributes we checked, what our confidence score is in that data, and have a feedback loop to our data scientists to continually improve the intelligence. That transparency is critical to our customers.

That’s a powerful capability. But does “AI agent reads the submission” risk sounding like you’re removing the human?

No, not at all, and this matters to me. These AI agents are augmenting intelligence, not replacing human judgement. The submission agent removes all the things the human had to do to get to a quote, but the underwriter is still there, leading the process and making the final decision. The whole point is to take the manual heavy lifting off people so they can do the judgment work. AI and humans manage the lifecycle together, and the human stays in the lead. That’s our governing principle: augment intelligence in the process, amplify the expert.

Let’s talk about what’s next. What is the CAIRA “orchestrator” and why does it matter?

Think of it as an AI manager: a coordination layer that knows what every agent can do, infers what a user is asking for, and calls the right capabilities to serve them. Take our Agentic MGA for the Cyber line. CAIRA, the orchestrator, infers from an inbound email that the user needs a quote indication. It reads the submission, pulls risk data from submitted documents and public sources where needed, infers the missing information, then invokes the Rater and Forms agents to produce a formal quote ready for underwriter’s review. Over the next 12 to 18 months, as that intelligence expands across policy, claims, billing, and distribution, this layer is where the leading platforms are headed. For us, that’s not CAIRA arriving, it’s CAIRA deepening.

There’s a tension in any deep technology: the capability is enormously complex, but the message has to be simple. How do you square that?

That’s the hardest thing we do, coming up with a simple story for a complex solution. But complexity and simplicity aren’t enemies. In engineering, we talk about encapsulating the core and wrapping functionality around it. The capability underneath can be massive, and the experience on top can still be clean, fast, and elegant. Look and feel really do matter, not just in the application, but in everything someone sees about us. The goal is to take a genuinely complex set of activities, simplify them, give them speed, and make them elegant and easy to use. That’s the opportunity, and that is what we are very good at.

As the industry continues to advance, what differentiates Cogitate?

The differentiation is in the depth, and in the fact that this is true agentic AI doing real work, not a label. It’s the difference between a statement on a website and intelligence doing real work the way a veteran underwriter or adjuster would begin it, trained to know which information matters out of hundreds of pages, and to hand the expert a decision-ready file. We don’t replace the thought process; we clear everything standing between the expert and the decision. Our agents are not just gathering information, but are also trained to know which information is important to pull out of hundreds of pages – whether that’s submissions, FNOL, or demand attorney language. And it’s the orchestration on top of that. Add the transparency we build in, the years of proven transactions underneath, and the speed of how we deliver, and that’s the package. Depth, made simple. The expert, amplified.

Last question. If you fast-forward, what does Cogitate look like, and what do you want the industry to understand about your direction?

The insurance policy lifecycle remains the same, end of story. What evolves is how intelligently the work to support it gets done. We’ll have one intelligence, CAIRA, working across policy, claims, billing, and distribution, orchestrated from the top, with humans firmly in the lead and AI clearing the friction underneath them. Over the last five years, we’ve reinvented ourselves; we’re several times the company we were. Today, we’re building for what insurance is going to need, not what it needed yesterday. We need to be one of the leaders in that shift. That’s the direction that excites me.