PM's and AI: Going Beyond the Easy Wins

Thanks for the questions, comments and all the messages on my hot take about AI: AI Hype Cycle. Whether they were in agreement, disagreement, or a mix, I enjoyed hearing how things are going with your product and AI. I got a lot of one question in particular: “How do you go beyond the easy wins and beat the curve on applying AI for your users?”

The general ‘how?’ is by 1) starting from what you know about the context of your UX’s to improve prompts

and

2) using that to invest your internal models with better and better proprietary data

A good PM understands the ‘Context’ of their UX better than chatGPT. That understanding should be your starting point for getting wins with AI (vs. the constant clamor about what cool trick AI’s just done).

For example, @khanacademy created an entire markup language for contextual prompt engineering to power their AI tutor, ‘Khanmigo’. So far, so good and it’s a platform for them to continually learn and improve.

For the big ‘beat the curve’ kind of wins, you’ll also want to continually run experiments to accrue data about what works for what users so you know more about how to leverage your user’s context than your competitors.

All the major platforms offer ‘enterprise’ plans where you can keep your data private. That part’s pretty easy. What’s hard, of course, is actually doing all this.?

If you don’t have good customer discovery, a healthy dev pipeline, and the capability (and motivation) to run experiments, AI won’t fix that for you. Are you out talking to and observing user behavior? Is your team focused on an economic definition of done, or just shipping features and bug fixes? Is your economic definition of done tied to clear, actionable target metrics? This is what Hypothesis-Driven Development is about. https://hdd.works/beatcurve