Scout InsurTech Rising with Kyber

Arvind Sontha is the Founder and CEO at Kyber, a Y Combinator-backed company building the AI-native platform for P&C insurance carriers to automate claims communications. Arvind was interviewed by Andrew Daniels to discuss Kyber's origin story, its rapid growth, and how autonomous back-office workflows could free insurance professionals to focus on what they do best.

Arvind, can you take us back to the beginning? What’s the origin story, and what was the moment you knew this had to exist?

“I left Google in May of 2022 to build Kyber. Originally, it was actually going to be an insurance company. I got my P&C broker’s license, got appointed by carriers, and started selling home and auto. But pretty quickly, I realized I wanted to leave the compliance, legal, and regulatory burden to the carriers that are best at it and focus instead on software, specifically, how to build an AI enablement layer that could make the industry an order of magnitude more efficient.

Insurance is built on text, documents, and unstructured data, which is exactly where generative AI thrives. We built a lot of early tools and workflows around that, but the real moment of conviction came when Charlie from Branch brought us a very specific use case. He’d been talking with his adjusters, and this was the number one thing they were complaining about: they were spending up to 40% of their time on tedious correspondence work.

Once we really dug into the workflow and everything required to get it right, it became clear there was an opportunity to build something 10x better than what existed in the market.”

What’s the specific gap or frustration Kyber was purpose-built to address, and why wasn’t anyone else solving it?

“One small but telling example is collaboration. Adjusters were drafting documents in Google Docs, collaborating there with managers, then copying everything into an enterprise system and spending hours reformatting it because the preview didn’t match the generated output. And in insurance, formatting matters; it has to match the policy language exactly.

Then, once the document is ready, the sending process gets messy fast. You’ve got multiple interested parties, different delivery methods, certified mail requirements, and questions about what can or can’t be sent electronically. At the end of the day, what the adjuster really wants is simple: hit send and trust that it’s handled correctly.

There were incumbents in the market, but they’d become so embedded in distribution networks that they didn’t really need to evolve the product for 20 years. A lot of those solutions are private-equity-owned, and when you open them up, they look like Windows 3.1. The space had been underserved for a long time.”

What does the workflow look like before and after, and what does that mean in real terms for cycle time, accuracy, or cost?

“There are really two workflows here: the adjuster workflow and the underlying operations workflow. Most people focus on the adjuster, but the operations side is a huge part of the pain.

With incumbent systems, getting even a single template live can take 20 to 40 hours, and that’s after training. The average P&C carrier has 100 to 150 templates just on the claims side. So you end up in situations where a carrier has been with an incumbent for a year, has maybe two to five templates set up, and nobody is really using the product because it’s too clunky. They fall back to Word documents.

One of the biggest things Kyber changes is that we have an AI template import tool that can take PDFs, Word docs, and custom formats and make them one-click publishable. That’s a big reason we’ve seen 43-day go-lives.

On the adjuster side, the workflow becomes much simpler. They trigger a document with a shortcut, the data is already pre-populated, jurisdictional clauses are selected, review cycles are enforced at the template level, and sending is one click. The user doesn’t have to think through who gets email, who gets certified mail, or how to route things correctly. The goal is to make it bulletproof and seamless for the adjuster, while also making it accessible and configurable for operations without needing IT involved.”

Since launching, how has Kyber grown? What’s driven the momentum, and where are you focused as you scale?

“This last year was a really big year. We’ve grown document volume about 100x over the last year, and we’re now doing well over a million documents annually. Revenue has grown dramatically along with that.

A big driver of that momentum has been trust. Insurance is a people-first industry. Our early development partners, like Branch, gave us the opportunity to prove we could solve a real problem and then extended their relationships and networks to help build trust across the broader market. Everything starts there.

As we scale, there are two major focus areas. The first is making sure we can maintain the same level of personalization, white-glove service, and customization for every customer as we continue to grow. The second is continuing to move upmarket. We see the pain most acutely in carriers north of about $100 million in gross written premium, and especially in the $500 million-plus range, where the coordination burden and operational complexity are much higher.”

How do you think about the future of this space? What are some of the problems that still need to be tackled?

“I’ve said before that the future of insurance is human. But I’d add that it’s also autonomous, and both of those things are true at the same time.

If you can make the back-office, regulatory, and process-heavy parts of insurance autonomous, then the people doing the work can become more human. They’re no longer buried in paperwork and tedious tasks. They can focus on what actually makes them great: analytical thinking, objectivity, empathy, and judgment.

Within claims, I think we’re going to see more of these high-volume use cases move into system-level processes through what we call flows. That means taking entire document workflows completely off the adjuster’s plate and having them triggered automatically by the core system, with record-keeping handled in the background.

I also think the same pain exists on the underwriting side. Today, underwriting teams still rely on Word docs, state filings, handoffs to IT, and long implementation cycles that can stretch into months. That process can be compressed dramatically and kept much more in underwriting’s hands, with IT only involved where it truly adds leverage. There’s still a lot of opportunity across insurance.”