Lenny Rachitsky: belief

Transcript context

…hoice still, I think, is a human element and needs that layer of additional judgment, which is very context specific to the company. So this is sort of right away, I think people should be doing this. I think more medium term and probably not too distant future and likely sooner than we all think, there's probably going to be, sort of the model that resonates with me is sort of the multi-agent model, which is you probably have different components of the strategy workflow automated. So you probably have a strategy agent, you probably have a roadmap or feature agent, you have maybe a engineering agent, and these can communicate amongst each other to cycle through results and iterate. And I think Armand Ruiz from IBM has some good definitional frameworks on some of the stuff. He shares it often on LinkedIn. But let's take a simple example. I can easily imagine something like this for a topic like onboarding. So every company and product team obsesses about their onboarding experience. And today, there are advanced experimentation frameworks. So imagine you're expecting a large surge in traffic. There's these sort of experimental frameworks like multi-armed bandits that can really help you get to the optimal sort of variation very quickly in real time. And there are variations of that, like contextual multi-armed bandits, there's combinatorial bandits. But the interesting thing is they still rely on human design of the variations, the different variations that you test. Even though the experimentation framework is very sophisticated, the variations are still human generated. Now imagine if those variations could actually be generated through generative AI and could be plugged into the advanced experimentation frameworks. The possibilities become infinite. And really, you might be surprised by what you find is the winning onboarding experience that you couldn't even have humanly imagined. And it might be different for every user, different for every sort of territory, etc. That is such a cool idea, I just want to say. This agent that's just running, thinking about ways to optimize your onboarding, coming up with concepts that you probably review, like, "Cool, let's try it." And then it does it, ships an experiment to run it and just is constantly optimizing your onboarding. Holy shit, that's an awesome idea. I think everyone will have this. And then what that makes me think there's going to be some company that's built the best onboarding agent. That's what you're going to be paying for. Totally. Totally.…

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