Kiriti Badam: prediction

Transcript context

…Okay. So let's actually follow that because that's a really important component of how you recommend people build AI stuff, AI products, AI agents, all the AI things. So give us an example of what you're talking about here, this idea of starting slow with agency and control and then moving up this rung. Yeah. For example, a very important or very prevalent application of AI agents is customer support. Imagine you are a company who has a lot of customer support tickets and why even imagine OpenAI is the exact same thing when we were launching products and there was a huge spike of support volume as we launched successful products like Image or GPT-5 and things like that. The kind of questions you get is different. The kind of problems that the customers bring to you is different. So it's not about just dumping all the list of help center articles that you have into the AI agent. You kind of understand what are the things that you can build. And so initially the first step of it would be something like you have your support agents, the human support agents, but you will be suggesting in terms of, okay, this is what the AI thinks that is the right thing to do. And then you get that feedback loop from the humans that, okay, this is actually a good suggestion for me in this particular case and this is a bad suggestion. And then you can go back and understand, okay, this is what the drawbacks are or this is where the blind spots are, and then how do I fix that? And once you get that, you can increase the autonomy to say that, okay, I don't need to suggest to the human. I'll actually show the answer directly to the customer. And then we can actually add more complexity in terms of, okay, I was only answering questions based on health center articles, but now let me add new functionality. I can actually issue refunds to the customers. I can actually raise feature requests with the engineering team and all of these things. So if you start with all of this on day one, it's incredibly hard to control the complexity. So we recommend building step by step and then increasing it. Awesome. And you have a visual actually that we'll share of what this looks like. But just to kind of mirror back what you're describing, this idea of start with high control, low agency, the example you gave is the support agents just kind of giving suggestions, is not able to do anything, the user is in charge. And then as that becomes useful and you are confident it's doing the right sort of work, you give it a little more agency and you kind of pull back on the control the user has. And then if that's starting to go well, then you give it more agency and the user needs less control to control it. Awesome.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.