Embedding AI Into Insurance Workflows: Where the Real Transformation Happens

That's great.

And I know that you've said that scaling AI is less about building better algorithms and more about actually embedding it into workflows like you just sort of discussed.

So can you talk about what that looks like in insurance and what sort of opportunities you're seeing there?

Yeah, absolutely.

We're at a point where AI is disruptive.

It's a massive change to all industries.

It's not an incremental improvement.

So you can always make better algorithms and make things move a little faster.

But if you fully embed AI into your workflows, that's where you're going to have that massive disruptive impact that produces this net new experiences for your customers.

We build with AI and our own workflows for developing code.

We use all sorts of different platforms for that, all of the major foundation models.

But we also use it for our workflows with our advisors.

The first one really is around education.

We want to make sure that we're educating consumers on the different types of policies they might be looking at.

As I mentioned at the beginning, I'm new at insurance and insurance is hard to understand.

I don't think it was ever built intentionally hard, but it's extremely difficult.

And so connecting all of these definitions and understanding all of that data allows us to build these educational experiences for our consumers and building that into the workflow is part of the natural point of it.

The next layer, like intake of the data.

So again, better algorithms.

We've had OCR for a while, object character recognition, and it's always gotten better at send me a photo of your driver's license or upload me a picture of your policy.

And we've been able to extract text for a long time, but now we're at this layer where we can pull so much context from those docs and make it available for our advisors.

So this intake and prep phase in the workflow is also extremely valuable there.

And then maybe the third place just to get a nice rule of threes here is in the guardrails, like creating agentic experiences, one of the difficult things for a long time was making it safe.

And so putting it into your workflows without knowing that it was going to work in all the corner cases was very difficult to do.

So in the same way that we use AI to build these workflows, we use AI to test these workflows.