Deep workflow understanding beats quick deployment

Most of the times, if you're obsessed with the problem itself and you understand your workflows very well, you will know how to improve your agents over time instead of just slapping an agent and assuming that it'll work from day one.

80% of so called AI engineers, AIPMs spend their time actually understanding their workflows very well. They're not building the fanciest and the most cool models or workflows around it. They're actually in the weeds understanding their customer's behavior and data.

I think we might've been in this weird temporary phase where, for a while, it was so hard to build product that you mostly just had to be really good at building product and it maybe didn't matter if you had an intimate understanding of a specific customer. But now I think we're getting to this point where actually if I could only choose one thing to understand, it would be really meaningful understanding of the problems that a certain customer has.

I often use this analogy of if you're doing a home renovation, you can have the most beautiful rendering of the new bedroom and we're going to have these lamps on the side of the bed that are coming out from the wall. But if you haven't checked if there's electricity in that wall there or not, it's going to drastically change the cost and the time and everything.

I think the first time we started growth, we could have been more user-centric and been a lot more hypothesis driven. We were following a lot of best practices that just didn't really apply to Dropbox.