I’ve come across a few posts now suggesting that AI is taking over the routine tasks of a product manager — things like writing specs, creating mockups, and handing off work from design to development — while the core responsibility remains: deciding which problem to tackle next.
I’ve never worked as a product manager, so — as you might have guessed — I naturally have an opinion on this! :)
In short, it seems to me that two different types of decisions are being conflated here.
The first type consists of decisions that are easy to reverse — for instance, which feature to build next, how to design the onboarding flow, or which screen to redesign. In the past, a mistake here could cost a month of design and development work, which is why you needed someone with a knack for making the right “guess.” Today, however, you can put together a rough version in a single day, launch it, observe the results, and roll it back if needed. You no longer have to guess; you can simply test.
The second type consists of decisions that are either irreversible or extremely costly to reverse. Almost all of these relate not to the product itself, but to the market, the target audience, and marketing.
For example:
I’m building Leaflo.app — a CBT-based journal for managing intense emotions. Here are the questions I spend a lot of time thinking about:
- Should it remain a journal, or should it evolve into a daily check-in tool?
- Should it be a simple, single-purpose app, or a collection of techniques for addressing various emotional challenges?
- Is it a personal diary, or a tool for collaboration with a therapist? And if it’s for use with a therapist, should that be a distinct marketing strategy targeting therapists, or just an optional feature?
- If I decide to create a mini-course, I’ll be entering a different league, facing new competitors with much larger budgets. Where would I find that audience? Can I actually reach them? Will I have enough money for marketing?
- And what if, this afternoon, OpenAI releases an AI-powered psychotherapist — what do I do then? What will psychotherapists and users actually do?
None of these questions can be answered by AI or by testing a rough, one-day prototype. If you get it wrong, you don’t just lose a day — you lose months and a significant amount of money. Just imagine having to rewrite all your landing pages to reflect a refreshed positioning strategy, or overhauling your SEO keyword strategy. It’s a total nightmare! :)
In my view, this is precisely where a product manager’s real work lies. AI has driven the cost of the first type of decision down to near zero. However, the second type of decision hasn’t become any cheaper, because the cost of a mistake here is determined by the market and long-term processes — not by how fast you can write code.
A good product manager knows how to distinguish between these two types of decisions. They won’t waste a week debating something that could be tested in a single day, nor will they rush into decisions that can’t be undone.
