Hamel Husain argues that generic "ask me anything" text boxes are a design failure because they set unrealistic expectations and give engineers no clear direction. Successful AI products need tightly scoped features built around specific user workflows, not all-purpose interfaces that try to do everything.
When I see a product with a little text box that says, ask me anything, I just know that we're in trouble right away. I know, I know that things are not going to be good. It's like, first of all, we're not communicating reasonable expectations to the user. What do you mean, ask me anything? You should never have that expectation be set. And I do encounter that quite a bit where people basically want to copy and paste ChatGPT into their product and say, well, you could just ask ChatGPT anything. You should be able to ask us anything and we can do anything. The people that haven't bridged that gap between like, well, okay, yeah, you know, that's different. It's not like you just copy and paste it, but you want to be able to do certain tasks in your application and want to be able to make sure those are like really good. I think it's really important from a product perspective to understand what your users want to do, what highest value use cases are, and try to scope it properly so that you can develop and test against something. It would be like telling an engineer, hey, build an app. And the engineer is like, to do what? And you said, I could just do anything. We wouldn't know what to build. Like no one can be successful in that regime, right? So it would be the same thing with AI. You don't want to do that.
Even if you look at your data too, you're like, oh, well, someone came to my app to buy groceries and then another person came here to journal. Another person came here to keep a spreadsheet. You're now going to build in 10 different directions. I think this is something that with, especially with voice, can be deceiving because the medium looks the same, the interaction model looks the same. And so features in voice are, you know, it's kind of hard to put your mind around like what is a feature in voice, but the feature is like a specific type of workflow. So it's, you know, what are you putting in that text box and what are you trying to achieve on the other side? Same goes for like when you're having a conversation, like what are you trying to achieve out of that conversation?
