"My PM is using AI to do my job. Is that good?"
—designers, engineers, marketers
I've noticed this question coming up a lot this spring. Fortunately, product teams have been here before.
Let me take you back to a time long before AI, about 15 years ago, when Mixpanel and Amplitude showed up. (Forgive me for reusing this analogy. I’m finding it so helpful in hard conversations about roles.)
[back to grandpa storytime voice]
Before self-serve analytics, if a PM wanted a retention analysis or a funnel chart, you filed a request with the data team, get in the queue. It sucked to be a PM and it sucked to be an analyst.
Then Mixpanel/Amplitude arrived, and suddenly I could make my own charts. I freaking loved this... aaand it handed me the rope to hang myself with.
I'd come to conclusions full of data fallacies. I'd misunderstand what an event actually meant, and miss gotchas that only someone who spent their career in data would catch. I didn't know what I didn't know.
Fortunately, I had data experts who patiently slapped me on the wrist until I finally learned my limits:
Low stakes: I completely self-served (these days, this might be a copy change AB test PR)
Medium stakes: I slacked a link/screenshot to gut-check my results (these days, this might be asking Claude to size a task)
High stakes, I’m calling a professional. To respect their time, I’d already done a few iterations and arrive at the meeting with a clear picture. (these days, this would be a working prototype before a kickoff)
That's how self-serve tools work. They shrink repetitive work, hand it to PMs (with guardrails), and free the experts for more interesting work. AI is just the latest incarnation.
That’s what happened to data teams. Before, they were up to their noses in data requests and maintaining dashboards. Now they were hired to think strategically and see the big picture across teams. Data roles are still very much needed, now for better reasons.
So, when PMs spin up designs and technical plans, the expert on their team is wondering: is this good? My answer is the same one that a data team would give about Mixpanel. “Yes, aaand...”
Practically speaking: PMs, don’t show up to your EM/Designer saying “look, Claude did it.” We don’t know what we don’t know.
Say, “I want to use AI to bother you less. How can I use AI to make your role better? When should I bother you?”
When I think about the best cross-functional relationships I’ve had, the expert owns the guardrails (and more strategic challenges) and PMs get to be more hands-on.
Neither of us wants to go back.