Grant Harvey

ChatGPT is amazing at solving 10,000 problems at once.

What enterprise solves 10,000 problems once?

No, enterprise solves one problem 10,000 times.

You have a very different set of expectations.

And there are different departments, different groups of people.

Often siloed, to your earlier point.

And so you've got to get agreement on what good looks like, on the right data that's coming, all these different things.

It's a very different challenge, and that's why you've got to approach it very systematically.

And the human is a key part of it in terms of understanding, in terms of valuation.

Honestly, I think it's weird that we're a company that spends so much time leading to these advances in these models and these products of these models, but actually all it's done is make us realize that our people are even more valuable.

Well, and that's not to say it's wrong.

There are going to be situations where that's hugely valid, but you've got to look at it coherently with clear eyes.

No.

How do you handle that?

Wait.

No, it depends on... Everything is a trade-off, right?

If you automate a process that's 30% correct, that could still be valuable, or it could be detractive because actually the people have examined the 30% more.

It's going to come down to cost, it's going to come down to speed, it's going to come down to the SLAs you care about.

And some people are fine with 80% accuracy on something because it's already higher than it was.

And we can just deploy something like that.