The new role of PM in the age of AI
The real competitive advantage lies in deciding what to build
For decades, the conventional wisdom in tech was that engineering capacity was the biggest bottleneck. Companies optimized around this assumption prioritizing ruthlessly to squeeze the most out of scarce and expensive development resources. But that assumption is starting to break down.
AI is Changing the Bottleneck
With the rise of AI-powered development tools, coding speed is no longer the constraint it once was. With the support of LLMs, engineers can generate, debug, and optimize code much faster. Tasks that used to take days now take hours. Entire workflows are being streamlined. While engineering is still the primary constraint, the trend is clear:
AI-empowered engineers can do more at a faster pace, releasing much of the tension that sets software development as a bottleneck.
Product managers can develop prototypes, workflows, and tools using AI-powered platforms like Cursor and Lovable in hours.
It is safe to assume that the trend will continue as these tools get better.
The Increase in Demand for Engineering
AI-empowered engineers can accomplish more, faster. As a result, the value of engineering is shifting. Repetitive, well-understood tasks are being commoditized, while activities that require expertise and creativity become even more valuable. Some argue this will reduce engineering demand, but I think we're witnessing another case of the recently trending for Jevons' paradox.
In case you're unfamiliar, Jevons' paradox states that increasing efficiency in a resource's use often leads to higher overall consumption rather than a reduction. Historically, when the steam engine improved coal efficiency, total coal consumption soared as more industries found it economically viable to use. The same principle applies to engineering efficiency: as AI makes development cheaper and faster, entirely new use cases emerge that would have previously been too costly to pursue.
For example, imagine a feature that provides a value of 100. If its development cost is 200, it wouldn’t be worth building. But if AI reduces the cost to 20 (a factor of 10x), it suddenly becomes viable. This increased efficiency unlocks previously unfeasible opportunities, leading to a rise in overall engineering demand.
The Changing Role of Product Managers
AI is also empowering PMs by enhancing their technical capabilities. In the past, PMs primarily focused on building a business case, conducting customer research, and aligning with stakeholders before a single line of code was written. Any market validation relied on interviews sometimes supported by mockups and static prototypes.
Now, PMs can validate ideas with functional prototypes - building lightweight versions of their products with API integrations, databases, and automation tools before engineering is even involved. Vibe coding apps enable PMs to execute more technical work independently, shifting how products are conceptualized, tested, and even brought to production.
Watch Lovable CEO Anton Osika developing an Airbnb clone in minutes. Ok this was a bit clickbaity. Anton is creating a clone of the Airbnb website look and feel but of course without any functionality, however he claims that someone with good prompting skills and some patience can develop the basic features to get a marketplace to life in far less time that this was assumed possible even a few months back.
These no-code tools on steroids remove barriers to entry, making everyone much more technical. What can we expect?
PMs will need to get familiar with these tools. When barriers to entry get lower, the table stakes rise.
PMs can be more useful in early stage teams. The answer to “when should a startup hire their first PM?” changes.
My thesis is that prompting is not a durable competitive advantage. Knowing how to prompt an LLM or a tool probably gives you a 3-6m head start until the next version of the same tool is good enough to deal with insufficient prompts. However, this is different. We are not talking about prompting for customer support queries or SEO optimized copy. This process (calling it prompting might be underselling it) includes several parts of the product work like clearly defining what to build and iterating with an agent to course correct. My prompting thesis will still apply for basic things like a portfolio website which AI is already really good at, but will give an edge to experienced users that want to build more complicated apps like an Airbnb clone.
What Changes?
As developing products is more accessible, we will have more1 product ideas, more experimentation, and more creativity. What will differentiate winning PMs (and their teams) are the insights on what to develop.
Does this mean traditional product artifacts like Amazon's working backwards approach will become obsolete? Some may argue that with AI-driven prototyping, extensive documentation will no longer be needed. However, I believe the opposite will happen: in a world where anyone can build, the real competitive advantage lies in deciding what to build (and then a killer execution). The winners will be those who excel at understanding customer needs deeply enough to build something truly valuable. The companies that win will be those that outthink, not just outbuild, the competition.
This is great news for founders who have a lower barrier to start a business as they can develop the first version of their product themselves and will have less pressure to raise money so they can hire expensive engineering resources.