💎 AIB - Executive Series - Dani Katz on how Natural-Language AI is changing Insurance Pricing Workflows

Hey it’s Fabio here,

In today’s Executive Series, I am joined by Dani Katz, Co-founder and Director at Optalitix, 

Across insurance, pricing transformation has been discussed for years, yet many firms still operate through fragmented processes and layers of spreadsheets. That raises a bigger question for the market:

What does it take to modernise pricing in a way that actually works for underwriters?

Dani has taken roughly 25 insurers through pricing transformation and has a clear view on why most of those projects stall. Dani explained why pricing transformation is an operational challenge, not simply a modelling problem, how AI can strengthen pricing and underwriting teams within existing structures, and why usability, transparency, and governance matter as much as technical sophistication.

Let’s dive in.

We've now taken approximately 25 insurers through a pricing transformation, and the same lesson keeps coming up: historic transformations fail because they're treated as a modelling problem, when they're actually an operational one.

Most insurers invest heavily in better modelling tools, analytics, and data but don't think enough about the underlying operational workflows: the underwriters, the governance, the deployment process, the business ownership that already exists. What happens next is predictable: underwriters look at the new system, decide they prefer the old way.

Underwriters' spreadsheets, however messy digitally, hold years of accumulated underwriting experience. They work well enough in a hard market - almost any pricing decision supports profitability there, even without real operational improvement - heading into a softer market, that's no longer good enough.

Replacing these tools is a change to the operating model. Our focus has always been on the operational side - building tools underwriters will actually use, embedding spreadsheets within our system, so the process works, before we make the pricing itself more sophisticated.

We’re genuinely excited about AI. What has changed is that newer AI models are now smart enough to work within existing insurance pricing processes and structures. That matters because pricing is regulated. AI needs to operate within the governance and controls of the structured pricing model, rather than replacing them.

In practice, underwriters can ask questions in natural language: “What are the key risks in this quote?”, “Show me similar risks and how they were priced” or “What is the concentration risk in this location?” They can also ask operational questions, such as which applications require approval. The system retrieves the relevant information and presents it in a structured way.

Effectively, you're giving underwriters access to the data and analytical power of an actuary or data scientist, without leaving the existing system. That's what's exciting about it — it lifts underwriters up, giving them tools and understanding they never had direct access to before.

This works because the AI pulls data straight from the pricing platform itself, using what's called Model Context Protocol (MCP) — broadly, a way for AI systems to integrate directly with existing software platforms, similar to how APIs let systems talk to each other. Not every client turns this on; some aren't ready to have AI inside the system, and engagement levels vary. 

But because it's pulling from the existing platform, every recommendation is traceable - the AI can't produce a price that sits outside the existing pricing model. It can only say "these are the inputs we think should feed into the model," or surface what the model is already saying. The pricing model itself remains the tool that ultimately sets the price.

This helps experienced people make faster, better-informed decisions. But the underlying pricing process must first be digitised. You cannot achieve the same result through a standalone spreadsheet; the data and models need to be connected through a platform.

There's often a false trade-off assumed between sophistication and usability. 

In our reinsurance systems, for example, actuaries handle around 80% of the pricing work - building simulations and complex experience analyses. They then produce a set of parameters for the price, which gets handed to underwriters. Underwriters see the part of the process they need to understand — the output — and they can adjust the assumptions to see how that changes the view. It's a sophisticated process underneath, but simple for the underwriter, because they only need to work within a defined set of parameters.

Effectively, we're presenting the results of sophisticated pricing models to underwriters in whichever medium they prefer - visualised in the browser, or pushed into a spreadsheet they can drill into. 

Both are possible within the same system, so you get sophisticated pricing with a genuinely usable interface.

And underwriters need to feel in control: able to change an assumption and immediately see what impact that has on the price. This is where natural language genuinely helps — because underwriters can ask a question in plain English, they're able to query complex pricing concepts that actuaries built, in a way that actually makes sense to them.

My view is that there'll be a complete reduction in reliance on manual processes — those get taken out of the underwriting system, and there'll be much clearer ownership and usability across pricing and underwriting. Underwriters and actuaries will still be very much involved, but they won't be doing the manual work they used to. Because that manual load goes away, the value they add shifts towards deeply understanding risk — getting into what actually causes it, and how to price for it better.

One thing that still surprises me about the P&C market is how often underwriting decisions get made without properly looking at the underlying claims and premium history that actually tells you more about the risk. I think that changes: every piece of data that can be analysed, will be. Manual processes will be gone, and all the relevant data will be instantly available to underwriters in a way that's easy to use.

The biggest differentiator will be the ability to modernise the entire process, from the pricing models all the way through to the underwriting decision itself. In a sense, underwriters will start making decisions more like actuaries do.

See you on Friday!

Fabio Caravita
Founder, AI Insurer Brief
[email protected]

Reach insurance execs and decision makers weekly.
72% Exec / C-level / Decision-makers.
Limited number of partner spots open each quarter.
→ Reply “Partner” for details and we’ll get in touch.

For Insurers, Brokers, MGA, consultancies and tech providers, exploring AI or with a live AI mandate but limited clarity on governance, processes, operating model design and execution.
→ Reply “Advisory” for details and the team will be in touch.