Will AI kill Product Management?
In this week's edition of Hiring Humans we have a guest blog post from Ravi Mehta (previously CPO at Tinder and EIR at Reforge) discussing how he sees product management evolving with AI.
If you’re short on time, read the 30 second version of this post.
In the last couple of years, product managers have been facing an existential crisis. It started with the tech layoffs and got worse when Airbnb eliminated the traditional PM role. Now, in the face of AI, Claire Vo predicts that product management is dead.
Too many PMs have asked me: Did I pick the wrong career? Will there be fewer jobs? Will opportunities be limited as every PM becomes a super IC? Will we even need PMs when products basically build themselves?
I see a different future for product management.
Yes, the role will change. But the traits that make a great product manager today will be as important (if not more) in the future: strong customer empathy, strategic thinking, and leadership skills.
To outline why, let’s think about product work in terms of its most basic elements: research and development.
Product Research focuses on gaining new knowledge through investigation and experimentation. It is a loop that starts by talking to customers, reviewing data to confirm a trend, and then running experiments to validate hypotheses. Research should focus on figuring out what to build.
Product Development takes the knowledge acquired from research and applies it to create a tangible product, process, or service that translates research findings into a practical application: coordinating the design and development of what a customer wants.
In simpler terms, research is about discovering new information, while development is about using that information to build something new or improved.
Today, AI has already changed product development. Each member of the product development triad (engineering, design, and product management) can move faster with the help of AI.
However, these productivity gains are not evenly distributed. A recent study found that generative AI improves engineering productivity by 20-50% for tasks like code documentation, code generation, test generation, and code refactoring. The entire engineering workflow is being accelerated by AI-amplified tools. In contrast, AI tools have had a more limited impact on product and design work. For reasons we’ll discuss below, this imbalance is likely to continue well into the future.
So what does that mean for the PM role? Many companies hire to a PM-to-engineer ratio. Often, 1 PM supports anywhere from 4 to 10 engineers. If those engineers deliver many times faster, we'll need more PMs to support those more productive engineers.
Think about driving a car. When you’re going slowly, changes to the steering wheel don’t do much. But, if you’re going MKBHD speed, tiny nudges can send you shooting off in the wrong direction.
The same is true for teams. As teams accelerate, the direction-setting that PMs do becomes even more important. I've seen this anecdotally when there are too few PMs: Teams spend much more time firing and less time aiming. They may initially feel they're moving faster, but if it's in the wrong direction, it's ultimately slowing down the rate of progress.
Figuring out this direction requires research work and involves:
Talking to customers.
Figuring out what they want.
Validating hypotheses based on these insights.
Does that get automated or delegated to AI? I don’t think so.
Predicting customer needs is hard. Before delegating a task to AI, we should consider how good we are at it in the first place, especially when that task requires intuitive insight. AI is good at replicating and synthesizing past decisions—it is trained on data that already exists. Making decisions about the future or predicting customer needs? Not so much.
A good PM has enough empathy not to take everything a customer says at face value and to go deeper when required. The process of figuring out what customers want is rate-limited by the customers themselves: we need to understand, analyze, and predict customer behaviour to generate and validate hypotheses.
That often requires observing what the customer does rather than what they say. A healthy level of scepticism is a good skill for a PM: watching for body language, noting a specific tone of voice, etc., and being alert to moments that require eye squinting.
Humans play a critical role in defining direction and decision–making, and the PM's role as a strategist, visionary, customer researcher, and analyst will not disappear. To do this well, however, requires a level of empathy that most humans struggle with, let alone AI.
Another point to mention when delegating roles to AI is accountability. As we discussed, steering is critical for fast moving teams. You'll want someone to put their credibility on the line to influence a team and ensure their ability to "aim" improves over time. Again, it is hard for even humans to do this well.
In its current form, AI gets us to the hard part faster. It helps us speed through mundane, easy work that previously required a lot of hours of "execution." But that rote work was only a small part of the job—which PMs have been trying to eliminate anyway. It will become much smaller in the future.
When we get to the hard part faster, we need PMs to push through that hard part—to focus on what customers want, define an accretive strategy, and coordinate AI-amplified people across an organization. Note how much of this requires human-to-human interactions. Yes, AI can help PMs, but it can't replace them (at least not yet!).
I don't see the future of product management in the hands of a small number of super ICs. Instead, we'll see product management take on an even more central role—and we'll see demand for what only the most senior and skilled PMs and product leaders can deliver today—deep product sense, rigorous strategic thinking, analytical decision-making, and the ability to build, lead, motivate, and align teams.
 | A guest post by
|