Which problem are you solving for?

Right.

So again, we've been around for quite some time and at this point we are a platform with offerings across basically every aspect of the full policy life cycle.

So that includes underwriting through to what has traditionally been called policy administration, through to claims, through to billing, and quite a few sort of small micro processes that sit in between.

But- I think the core ethos of what we have always been focused on is this idea that it's well understood that in insurance, no one risk matters.

It is entirely about portfolio construction.

The very idea of insurance is that these risks balance each other off.

What were those kind of lessons when you got your first sort of hard job, hard-paying job in the insurance industry? What did you learn from that?

I guess

And do you think those challenges are purely technical or do you think they come from more of a talent and mindset gap or is it a little bit of both?

You know, it's so funny.

I think there's a little bit of both for sure.

I just sent an email just a few minutes ago to a senior leader in an insurance company.

And it was about someone that we both know who's coming right out of his undergraduate.

And we were just sort of going back and forth on, you know people are slow to hire young folks coming out of school right now because they think oh ai is going to scale my more experienced labor but we were both looking at this individual and we were looking at the sorts of school projects they had sent us what's called a jupiter notebook basically a large python uh code uh file uh showing um a lstm based uh transformer model.

So I think like taking an LLM and building like a predictive model, using that LLM to create a sort of a vector representation of some underlying data.

I guess

And do you think those challenges are purely technical or do you think they come from more of a talent and mindset gap or is it a little bit of both?

You know, it's so funny.

I think there's a little bit of both for sure.

I just sent an email just a few minutes ago to a senior leader in an insurance company.

And it was about someone that we both know who's coming right out of his undergraduate.

And we were just sort of going back and forth on, you know people are slow to hire young folks coming out of school right now because they think oh ai is going to scale my more experienced labor but we were both looking at this individual and we were looking at the sorts of school projects they had sent us what's called a jupiter notebook basically a large python uh code uh file uh showing um a lstm based uh transformer model.

So I think like taking an LLM and building like a predictive model, using that LLM to create a sort of a vector representation of some underlying data.

So can you talk about the importance of the data that's used for AI and some of the best strategies to make sure you're training your models with clean data?

Yeah, sure.

So, I mean, I think it's important that last thing you said, which is training models is, you know, yes, if you are building models and training them, clean data is table stakes.

At the same time, so much of what the current AI evolution is about is that you don't have to train models anymore.

Right.

The point of a GPT, right, the P stands for pre-trained.

And so the point of a GPT is that it's a generative pre-trained transformer, meaning that you are able to take the fact that this model was trained on really the English language holistically and has a very broad-based understanding.

But I'm curious over the next couple of years, what will separate the insurers that are winning with AI from those that are sort of falling behind?

So can you talk about the importance of the data that's used for AI and some of the best strategies to make sure you're training your models with clean data?

Yeah, sure.

So, I mean, I think it's important that last thing you said, which is training models is, you know, yes, if you are building models and training them, clean data is table stakes.

At the same time, so much of what the current AI evolution is about is that you don't have to train models anymore.

Right.

The point of a GPT, right, the P stands for pre-trained.

And so the point of a GPT is that it's a generative pre-trained transformer, meaning that you are able to take the fact that this model was trained on really the English language holistically and has a very broad-based understanding.

But I'm curious over the next couple of years, what will separate the insurers that are winning with AI from those that are sort of falling behind?