Keri Briske

How will you verify it?

And so once you get data, once you have your evaluation metrics and your criteria and your use case, then you can start to expand your use cases and expand your accuracy and start to grow your model needs.

especially your specialization needs.

We see a lot of customers, if you don't mind me to kind of digress for a second.

No, sure.

It's kind of like these eras since the big bang of chat TPT in 2022, right?

Enterprises were like, oh, what is generative AI?

What do I do with it?

Like, how do I use it?

Like, I need something similar to that.

And they're like, oh, well, I actually don't have the resources to train my own model, so I'm just going to use the managed APIs.

And then everyone kind of went on that proof of concept.

And then in 2023 was all about that.

And 2024 was about, if you've heard of RAG, Retrieval Augmented Generation, they're like, oh, you know, models hallucinate.

Well, now we need RAG.

RAG's my silver bullet.

Like I'm going to, all I need is RAG.

And so they started testing with RAG and they realized it's just, you know, you kind of can only do so much with it.

And and then they're like, oh, you know, I might need to customize my model.

So they came back to the conversations that we had in the beginning of 2023.