Is AI Rewriting Product Strategy? Not the Way You Think

The short answer: no, and that's the interesting part

AI has not rewritten product strategy. It has rewritten the cost structure underneath product strategy, and that is a more disruptive change than a new framework would be. The questions that define strategy, who you serve, what problem you solve, why you win, why now, are exactly what they were in 2019. What changed is that execution, the thing most product organizations were actually built around, stopped being scarce.

That inversion is uncomfortable. For two decades, the constraint in software was building. Roadmaps, sprint rituals, headcount plans, the entire operating model of a product org existed to ration engineering capacity. When a quarter of Y Combinator's Winter 2025 batch shipped products with 95% AI-generated code, and companies reached $10 million in revenue with teams of under ten people, that constraint quietly dissolved. The bottleneck moved upstream, to the quality of the decisions about what to build at all.

What actually changed: execution collapsed in price

Garry Tan reported that the entire YC Winter 2025 cohort grew roughly 10% week over week, something he said had never happened in early-stage venture, and attributed it directly to AI taking over the build workload. Founders no longer need 50 engineers to test a thesis. The capital lasts longer because the construction phase shrank.

Inside larger companies, the same compression shows up function by function. McKinsey's State of AI survey found cost reductions concentrated in software engineering and IT, with revenue gains most commonly reported in marketing, sales, and product development. The grunt work of product management, synthesizing feedback, drafting PRDs, summarizing research, is being absorbed the same way the grunt work of coding was.

So far, so familiar. This is the part every "AI is transforming PM" post covers. The part that gets skipped is what cheap execution does to strategy.

What didn't change: the strategy questions still have to be answered by someone

A fact worth sitting with: 88% of organizations now use AI in at least one function, per McKinsey, yet only about 6% qualify as high performers attributing 5% or more of EBIT to it. MIT's GenAI Divide report found 95% of enterprise AI pilots delivered no measurable P&L impact despite $30 to 40 billion in spending. The lead author's diagnosis was blunt: not model quality, but a learning gap in how organizations integrate the technology.

Read those two studies together and the picture sharpens. Almost everyone adopted the capability. Almost no one converted it into business outcomes. That is not a technology gap. That is a strategy gap wearing a technology costume.

The MIT data makes the misallocation concrete: over half of GenAI budgets went to sales and marketing pilots, high visibility and low return, while the measurable ROI sat in unglamorous back-office automation. Companies didn't lack AI. They lacked a defensible answer to "where does this create value for us specifically," which is the oldest strategy question there is.

The new failure mode: velocity without direction

Here is the inference I'd defend: AI doesn't make product strategy better or worse. It makes strategy higher leverage in both directions. When shipping a feature took two quarters, a mediocre strategy produced slow mediocrity, and you had time to notice and correct. When shipping takes two days, the same mediocre strategy produces a flood of well-built, irrelevant features before the next quarterly review.

McKinsey's high performers illustrate the inverse. They are 3.6 times more likely to pursue transformational change rather than incremental automation, and the majority fundamentally redesign workflows when deploying AI rather than bolting it onto existing process. They set growth and innovation objectives, not just efficiency. In other words, they did strategy work first and let AI accelerate a direction they had already chosen. The other 94% accelerated whatever direction they happened to be pointing in.

I see this pattern weekly reviewing AI product work: teams arrive with impressive demos and no answer to "why would the incumbent not do this next quarter for free." The demo got cheap. The answer didn't.

Where the moats moved

If anyone can build your feature in a weekend, the feature is not the moat. This was always technically true; AI made it operationally true. Defensibility has shifted to the things that cannot be vibe-coded:

• Distribution. Reaching the customer is now harder than building the product. The order of difficulty flipped.
• Proprietary data loops. Products that learn from usage compound in a way clones starting from zero cannot. MIT's research found the winning 5% of deployments were exactly these: systems that learn from and adapt to workflows, versus static tools that don't.
• Workflow depth and integration. Being embedded in how a customer actually operates, with their data, permissions, and edge cases, is expensive to displace even when the surface UI is trivial to copy.
• Judgment and taste. When everyone has the same model access, the residual differentiator is the quality of choices about what to build and what to refuse. That is a strategy asset, held by people.

Speculation, flagged as such: I expect the next two years to be brutal for products whose entire identity is "X, but with AI," because the marginal cost of producing another one is approaching zero. The survivors will be the ones whose AI sits inside a loop that gets harder to leave every month.

What this means for how you operate

Three practical shifts, in order of importance.

Match strategy cadence to execution cadence. Annual strategy reviews made sense when building took quarters. If your team can ship in days, a strategy reviewed yearly is stale for 11 months. Revisit positioning and bet allocation monthly. The document can be short. The cycle cannot be long.

Treat the roadmap as a portfolio of reversible bets, not a delivery contract. Cheap execution means the cost of being wrong dropped, but only if you actually kill losers fast. Most orgs ship faster and kill at the same old speed, which just accumulates well-built clutter. Pair every accelerated build with an explicit kill criterion.

Reinvest the reclaimed hours deliberately. AI handing PMs back their synthesis time is only a win if that time goes into the work AI can't do: talking to customers, sharpening positioning, making the uncomfortable focus decisions. If it gets reabsorbed into more stakeholder meetings, you've automated the wrong half of the job.

Conclusion

So, is AI rewriting product strategy? The principles, no. The stakes, completely. Execution used to hide weak strategy; a slow build cycle gave everyone plausible cover. That cover is gone. The 88%-adoption, 6%-impact gap is what strategy debt looks like when execution stops absorbing the blame.

The teams that win from here won't be the ones using AI most aggressively. They'll be the ones whose strategy layer is sharp enough to deserve the speed.