The AI reforms throughout the tech sector are reshaping how we build products and will continue to do so. Somewhere changes are slow, somewhere they are faster (usually, in smaller teams with less security and regulation-imposed guardrails), somewhere the redemption arc of banning or rejecting AI is already happening. What matters is that it will stay with us, spread across evenly and impacting products we use in our daily life and at work.
I am not an apologist for any of those changes, and mostly fact-checking them in my head against how I see my work as a product manager. And while the LLMs can speed up our discovery and delivery cycles, I see the exact issue here why product management will remain important.
We did instill ourselves the mantra of moving fast and breaking things. Not only when you are small enough to do so without huge collateral damage, but also when you are big enough to crash half of the internet, or damage the teens’ psyche for the means of a higher engagement, or just making the user experience worse with numerous A/B testings tied up with metrics standing way too far away from the end user. You can find related cases in the recent history of the big tech companies for yourself. And still, there is a limit of this much people can do to maintain this pace without degrading the quality of their decisions.
If you discover and implement features and other changes faster and faster, your ability to see what does good and what doesn’t blurs with the same speed. And any feature either does one or the other in overall consensus, weighing all pros and cons.
Seeing how we do now, we can see we are not very good at it even now. How often would you see products remove features from the product instead of just abandoning them? How often do our teams just release and release and release, looking only for success and not providing proper action on failures? Where “less is more” is a thing still?
Nobody even talks about end users who are the ones who need to adopt these changes, to incorporate them into their mental models, which is on its own a very limited capacity of the human brain. I, as an end user of many products, do not want to see them adding and removing dozens of features per month for the sake of, in this case, only improving metrics related to my experience only by a long stretch. At such speed, users won’t even experience the goodness of the changes as they will be stripped away and replaced by other changes. Or if not, the product will be bloated to such an unusable extent that anything good will be a minuscule part of it.
I can imagine the world where product managers per se are not needed. Only because the developers, or designers will become ones. Or vice versa. But the skill set of defining the product vision, what’s good and what isn’t, will stay essential and will be even more important to acquire. The ability to defend end users is something an LLM can’t do at all (it’s funny even to imagine how it does it). The systems thinking around does imply to think about all those parts surrounding the product, not only the product itself. And someone should balance those parts of the system.
These points above are also good ones for any PMs out there who do not have any hard skills. Learn the developers’ argot at least. Become better at designing UX on different levels, from connecting jobs-to-be-done scenarios together and nesting them to the very buttons and other UI elements. Be good with defining and analyzing metrics in different tools. LLMs are not magic machines, they are probabilistic by nature, meaning they provide you the average, not the best.
The note was written during my vacation. The spring sun has already started to pave the way for the grey country I live in into warmth. Still, some snow piles are lying here and there, covered in city dust, postponing the inevitable. It’s fine. They’ll go too soon enough.