Every product manager I work with right now is using AI to move faster. Specs get drafted in minutes, competitive scans happen before coffee, and roadmap slides practically build themselves.
Almost none of them have noticed what they’re trading away to get that speed.
If your AI tool can generate your positioning, your competitor’s AI tool can generate theirs too. Same training data, same patterns, same instinct to reach for the most statistically likely answer, which by definition is the most average one. You haven’t found an edge. You’ve found the same starting point as everyone else in your category, including the private label manufacturer who’s about to undercut you on price with a product that’s “good enough.”
This is the textbook definition of commoditization I’ve been teaching product managers for thirty years: standardization, low barriers to entry, increasing price transparency, products becoming indistinguishable from one another. We used to talk about this as something that happened to an industry over the course of a product life cycle. Now it can happen to a single product launch in a single afternoon, because the tool that was supposed to make you faster also made you generic.
AI Is Brilliant at the Science. It Has No Idea About the Art.
I’ve spent my career telling product managers the same thing: your job is the art and the science of persuasion. The science is data: customer lifetime value, retention rates, pricing analytics, and market sizing. The art is everything else: reading a stakeholder who hasn’t said what they’re actually worried about, finding the unmet need a customer can’t quite articulate, building a narrative that connects a product feature to a CFO’s anxiety about next quarter.
AI has gotten remarkably good at the science. Feed it a CLV equation, and it’ll calculate retention-weighted margin in your sleep. Ask it to summarize ten competitor pricing pages, and it’ll have a table back before you’ve finished your coffee.
It has no functioning instinct for the art. It cannot sit across from a frustrated customer and notice the thing they didn’t say. It cannot read a room of skeptical stakeholders and adjust the pitch in real time. It cannot walk a value chain, talk to the unhappy customer, watch the distributor get squeezed on margin, notice the white space a competitor’s roadmap accidentally left open, and turn that walk into a story that makes a skeptical VP lean forward. That walk has always been the job; AI has simply made it obvious by automating everything around it.
Observation, instinct, and empathy are not soft alternatives to data. They are data, the kind that doesn’t show up in a dashboard until long after a competitor has already acted on it. A product manager who notices a customer’s hesitation before it becomes a churn statistic or senses a stakeholder’s unspoken objection before it kills a project in committee is running analysis just as real as a CLV model. It’s simply an analysis of a model that can’t run for you, because the inputs only exist in a room, in a conversation, in a moment of attention that a transcript doesn’t capture.
The Shortcut That Costs You the Thing You Were Looking For
Under deadline pressure, the instinct is to let AI shortcut the art too, not just the data crunching, but the strategy, the narrative, the “what’s our unique angle here.” It feels efficient. It is, in fact, the fastest route to the exact outcome you’re trying to avoid.
Differentiation has never come from the parts of the job that are easy to automate. It comes from the parts that are slow, a little uncomfortable, and resistant to shortcuts: sitting with a customer’s actual frustration instead of a survey summary of it, walking the full value chain instead of pulling a competitive matrix off a website, sitting with a stakeholder’s real, often unstated, objection instead of drafting around it. A product manager who skips that walk because AI handed them a polished spec was never going to find anything the competition’s AI couldn’t also find. They’ve optimized themselves straight into sameness and handed private label exactly the opening it needed.
Private label isn’t winning right now because retailers have gotten smarter. It’s winning because too many branded products stopped explaining, in a way, a real customer would actually feel, why they’re worth the difference. AI can’t fix that gap, and, used carelessly, it widens it because “good enough, fast” is precisely what a private-label competitor is already optimized to deliver at a lower price point. You cannot out-AI a private label manufacturer on speed and cost. You can only out-narrate them on value.
This is where innovation actually lives in the value chain: not at the end, as a feature bolted onto a finished product, but at the point where the walk turns into insight. A product-centric organization asks what it can build next. A customer-centric one asks what experience it’s actually delivering, because the product is the experience, and the experience is the product. Customers don’t experience your spec sheet. They experience how it felt to discover the problem you solved, to buy the solution, to use it, and to call support when something went wrong. Every one of those moments is a place innovation can show up, not just in the product itself, but in how it’s sold, serviced, and remembered. The product manager who walks the full chain finds those moments. The one who outsources the walk to a model never sees them at all.
The Walk Is a Discipline, not a Metaphor.
So, what do you actually do differently this week?
Start treating the value chain walk as a deliberate exercise rather than something that happens to you when you have spare time, because under deadline pressure, spare time never arrives on its own. The goal isn’t just to find the friction. It’s to turn what you observe into value the customer actually feels, which is where real innovation starts, long before anyone opens a spec template.
- Find the customer who’s irritated, not the one who’s satisfied. Satisfied customers confirm what you already know. Irritated one’s hand, you the unmet need AI can only guess at from a review summary, and the unmet need is where innovation actually begins.
- Stand where your channel partner gets squeezed. Distributors, retail buyers, and dealers, whoever sits between you and the end customer, absorb pressure that your own data won’t show you. That pressure point is usually where the real opportunity is hiding.
- Read the competitor’s roadmap for what’s missing, not what’s there. AI is excellent at summarizing what a competitor is doing. It’s far weaker at noticing the segment, or the experience, they’ve quietly abandoned.
- Write the narrative before you write the spec. If you can’t explain, in plain language, why a real person should care, you don’t have a product yet. You have a feature with a price tag.
None of these four things can be outsourced to a model. Each one takes the same hour AI just bought you back from drafting a market summary and reinvests it in the only part of the job that still belongs to you.
This is precisely the muscle we built the Consumer Products Institute of Canada’s commercial training programmes to develop deliberately in young product management professionals, rather than leaving them to stumble into it by accident a decade into their careers. The technical side of product management, the specs, the data, and the process can increasingly be learned from a tool. The art of persuasion, stakeholder empathy, and value-chain thinking has to be coached, practiced, and built through real product work with real customers. That’s the differentiator: CPIOC exists to build early, before a career’s worth of bad habits sets in.
What This Actually Changes About the Job
This isn’t an argument against using AI. I’d be a hypocrite and a bad teacher if I told you to ignore a tool that genuinely makes the science half of your job faster and better. Use it for the CLV models, the market sizing, the first-draft competitive scan, and the performance analytics dashboard. That’s real time saved and real rigor gained.
The argument is about where you spend the time it frees up. If AI buys you back six hours a week, the only acceptable use of those six hours is the four things above: not more dashboards, not another AI-assisted brainstorm, but real conversations with real people that a model has no access to.
The product managers who thrive over the next five years won’t be the ones who used AI the most. They’ll be the ones who used it to clear away the science, so they had more room for the art, and who understood that the art was never a soft skill. It was always the actual job, and AI has simply stripped away everything else, leaving the one thing that was hard enough to matter in the first place.
The question worth asking yourself this week isn’t how to get AI to do more of your job. It’s whether, now that AI can handle the easy half, you’re actually doing the hard half, or whether you’ve just found a faster way to be forgettable.