When you hear all of that, do you think risk? Do you think things that you have to overcome operationally to really consider that journey to get to that?

Yeah, I think it depends what you're using it for, right? I think in something like product discovery, right, it's fairly low risk in the sense that the humans doing the review of what the AI research is producing and then ultimately making the decision on what product or service to buy.

I think where the risk comes into it, particularly in financial services, is if you start automating certain things like fully using agents or AI, you have to have a very high degree of confidence that it's going to do what you expect it to do.

So I think that's what I'm hearing from policymakers and regulators and other folks in the industry is the capabilities are amazing, but how do you clearly say to the regulator or say to your customer that what's going on in the back end with the agent is what you expect it to do? So everyone uses that term hallucination.

Hallucination is not maybe that risky when you're looking at what plumber to hire or what product to buy for your 13-year-old niece.

But it's super risky if you're doing automated customer onboarding or like investigating fraud or transaction monitoring or something, right? So you got to think about what it's actually being used for and the risk profile is quite different.

And thinking about what you mentioned earlier from a treasury perspective and thinking about moving a trillion dollars, which I wish I had in my bank account.

Managing liquidity for a global payments company at that scale is complicated, right? We've got thousands of bank accounts and hundreds of currencies in lots of different geographies across different products and services.