What we get wrong about being “technical” as a product manager
Why AI is still far from replacing product managers
Years ago, I was a product member in a 10-person AI start-up. Half of the company was made up of AI research engineers. It was a super smart, sharp, fast environment that I was really proud to be in. I also had an engineering degree, which I had graduated with three years earlier. Although it was mostly focused on statistics, I still had some coding experience. But when it came to the domain itself, machine learning know-how, I started from zero.
Funny how it’s common to hear “AI company” or “AI-based product” now. Back then, I didn’t personally know anyone else working with any type of AI, not even integrations, let alone training their own models. I remember being overwhelmed in the beginning and feeling very incompetent during meetings where we had to talk about results or solutions.
Then the pandemic came. Perfect timing. Staying home all day, all week, I finally had time to learn more. I was working more than 12 hours a day during that period, and I enrolled in online machine learning courses, watching classes at 6am and 11:30pm, yes. Deep learning with Keras, NLP in Python, tree-based models, TensorFlow… In a few months, I finished them all and completed real projects successfully using Python to get my certificates.
I remember nothing.
Well, not as in I don’t understand how these concepts work or what different approaches can be used for certain problems. I still know those today, if not even better. But I don’t remember how I completed those projects, how to actually execute them, or what I would do now if I were given those datasets again.
Even though I don’t regret doing it, I personally think the best way to understand something is to get your hands dirty, I have to admit it wasn’t the smartest move.
Fast forward to today, where “AI” is used by everyone, even though most of the time they refer to LLM-based AI assistants (ChatGPT, Gemini, Claude, etc.) or AI-powered coding environments and product builders (Replit, Lovable, etc.). We can now do in hours what I was trying to do in months six years ago. Product managers can sketch, make prototypes, write scripts, or even build things in hours.
On social media, there are a lot of arguments around this. “Do we really need product managers to become junior engineers or designers by using AI? Shouldn’t they focus on what they normally do instead of shipping low quality engineering work through vibe coding?” I don’t find this a smart argument. If it takes 10 minutes for a product manager to build a prototype to share with engineers, instead of writing a long spec and having a two-hour meeting to explain it, leave them alone and let them do it.
But is this the most important benefit of AI in product management?
I was a product person working in AI, trying to learn the engineering behind it without AI assistants available. Now I’m a product person working in a different domain with multiple AI assistants available. I think I can share my two cents on this:
The biggest change AI brought to my PM work is where I spend my thinking energy.
The improvement curve is exponential. And even next year, what we’ll be capable of doing with AI will probably be unbelievable compared to today. I rarely think about technical limitations now compared to the past. Knowing that technology will follow at some point, I make stronger decisions about what needs to be done rather than fitting ideas into current technical possibilities. I have never felt more empowered and independent in my life as a product person.
While years ago I felt the need to deeply understand tools and how to use them, now I don’t invest heavily in mastering any of them unless they’re core to what I’m doing. I still strongly believe that having skills in at least one or two of design, engineering, or data is a must-have asset in product management. Today, though, it’s more about being able to ask the right questions.
So yes, a product manager must be able to frame a hypothesis and define what success metrics will prove the version they’re testing is successful. But it’s no longer about writing the perfect query to retrieve them, because AI assistants can help whenever they’re stuck.
Similarly, a product manager still needs to understand user behavior. They need to be able to ask questions like which flow reduces churn or where users might get confused. What they no longer need is to describe an experience purely in text. A PM can now use AI to generate a rough flow or interface representation in minutes. With this thinking artifact, designers already have a starting point. It’s up to them how much they pivot from there, but it saves a lot of time and confusion.
Instead of having hypothetical conversations with engineers, product managers can now use AI to sketch a simple logic flow or pseudo-code to explore how an idea might work, even if they’re not technical. It’s not the final solution, and it doesn’t have to be. But it lets them identify bottlenecks, constraints, dead ends, or gaps in logic much faster.
In the story I shared at the beginning, I was chasing execution skills. Today, I believe the skill that matters most is judgment. PMs aren’t becoming engineers or designers, they’re becoming faster at deciding what not to build. This is an era where product managers have never been able to do more product work, and I can’t wait to see what’s next.