Stop Waiting for New AI Models: Here's How to Use Data to Make AI at Work, WORK

We need to stop thinking that AI is going to get better at work (or at home) if the models get better. It DOES help (a bit), but the real value comes from getting the data right (glamorous I know)

Look this is just straight up a spicy take and lots of people will disagree.

Everyone wants to ooh and ahh about the latest model. The model makers themselves encourage you to think like this. Wait and the additional intelligence will fix everything!

Next month’s models will matter! o3 Pro coming out will change it! (That is probably coming this week by the way.) Great. Fine. I’m not here to say intelligence scaling doesn’t matter—if you read this Substack you know I take the search for a new everyday model very seriously.

But, my observation is that the best model in careless hands is worse than useless. Work doesn’t get done faster. Rewriting customer emails to make them 10% more professional isn’t really adding value at this point.

The value is in actually thinking about how we handle data inputs to AI, but that takes work and to be honest it’s not as popular. If we took 50% of the passion we put into the latest model and put it into getting our data inputs right we’d get twice the leverage. Or more. So yes, prompts matter (I’m working on that here). But so does your actual data! And that’s what this post is about. How can you practical use your data with AI?

I give you a real life example, and some principles to think about, and I call out where I think this highlights the need for more human judgement, not less. Enjoy!