Product Operations in the age of AI

AI has broken Product Ops' standardization playbook: workflows now expire faster than they can be rolled out. The new mandate is making change cheap, workflows as products, composable blocks instead of monoliths, time-to-adopt as a metric, reclaimed hours reinvested on purpose, while investing in the one asset that appreciates: taste, trained against baselines. Organizations currently pour money into the depreciating asset and ignore the appreciating one.

A year ago I wrote a careful document about AI tools for product teams. It was part inventory, part gentle argument: the enablement work our craft leads were already doing deserved a name, and the name was Product Operations. It had a comparison table: capabilities, pricing, which tool fit which phase of work. It was accurate when I wrote it. Most of the table is already out of date. That is not a failure of the document. It is the most important fact about running a product organization right now.

Product Operations grew up as a standardization discipline. The function exists to reduce friction, improve decision quality, and let product teams scale without chaos: shared templates, shared tooling, shared reporting, shared ways of working. That mission still stands. The method behind it is what's breaking.

The old method assumed workflows were durable. You could spend a quarter rolling out a template or a Jira configuration because it would stay true for years. That assumption is gone. Productboard's 2025 survey of 379 enterprise product professionals found that every product team surveyed now uses AI, most PMs daily, saving around four hours per task. 88% of teams run two or more different models, with no stable market leader. ¹ The ground under every workflow is moving, and it is moving faster than any playbook can be written. Sherwin Wu, who leads engineering for OpenAI's API platform, put it plainly on Lenny's Podcast: "models will eat your scaffolding for breakfast." ³

So the job changes. The old Product Ops mandate was to standardize the best way of working. The new mandate is to make changing the way of working cheap. In practice, that looks like this:

Treat workflows as products. Version them, measure their adoption, and give them sunset dates. A workflow without a retirement plan is future operational debt. The tool table in my old document didn't need to be more accurate. It needed an expiry date and an owner.

Build blocks, not monoliths. A pitch needs radically different support than a long-term delivery engagement, and that was true before AI. What AI adds is speed of change, which punishes any process rolled out as one rigid system. The teams that adapt fastest are the ones whose operating model is a menu of small, composable practices that each PM can assemble for their context.

Measure time-to-adopt, not just time saved. The interesting gap is no longer who uses AI. ProductPlan's 2026 State of Product Management report found most teams parked in limited workflows or early experimentation, with only 6% treating AI as a core capability: usage is near universal, depth is rare. ² The separation happens at the extremes, between teams that absorb a new capability in weeks and teams that take quarters. Time-to-adopt is a measurable, improvable operations metric. Almost nobody tracks it.

Reinvest the reclaimed hours on purpose. If AI saves your PMs four hours per task and nobody decides where those hours go, they evaporate into more throughput. The 2026 data shows exactly this default: teams widely report time saved and faster synthesis, while only 11.5% report more confident prioritization. ² The saving only matters if it buys judgment time: strategy, and the closer client partnership that repetitive delivery work crowds out. That reallocation is an operations decision, and defaulting on it means automating faster in whatever direction you were already pointed.

None of this is theoretical for me. One of my favorite recent assignments was joining an internal rapid-prototyping team: AI strategists, tech leads, creative leads, and me, working out how to productize the craft of rapid prototyping itself. A lean team of experts wearing multiple hats and building together, treating the way of working as the product. The questions we were answering were operational ones: how much time, which experts, what process, which flows, so the offer could be scoped confidently across a growing number of pitches and vision engagements. It settled something for me: reimagining how teams work is the opportunity AI brings, not the tax it imposes.

Which brings me to the part that doesn't depreciate. When drafting a PRD takes minutes, PRDs stop being a differentiator. When anyone can prototype, prototypes stop being a differentiator. What's left is what was always the real job underneath the artifacts: knowing which problem is worth solving and recognizing when a fluent output is wrong. The industry has converged on a word for this. Tony Fadell spent an entire recent Lenny's Podcast episode on building taste and judgment in the AI era, ⁴ and OpenAI's Codex product lead Andrew Ambrosino describes taste as the most valuable professional capability in an AI-first workplace. ⁵ Productboard's data points the same way: teams use AI overwhelmingly for synthesis and sense-making, while prioritization and problem selection remain human. ¹

Taste is not mystique, though. It gets trained, and it gets trained against baselines. The parts of my old document that aged best were not the tools. They were the templates and playbooks, because they encode a standard: a working definition of what good looks like. Every new AI workflow I have built since was measured against that baseline. Fast and cheap are easy to buy right now. Quality has to be defended, and the old artifacts are how you defend it.

And here is the gap that worries me. Productboard's survey of product leaders found 85% investing in AI tools while only 2% prioritize talent development, and PMs agree: lack of training is one of the top two blockers they name to getting more value from AI. ¹ That is the whole problem in two numbers. Organizations are pouring money into the depreciating asset and ignoring the appreciating one. Or as Setu Shah, Senior Director of Global Product Strategy at Oracle, put it in the 2026 State of Product Management report: "AI will be easy to buy. Judgment will be hard to scale." ²

Great product managers were never great because of their tools. The tools used to hide that. Now they expose it. Product Operations in the age of AI has one job with two halves: build the system that makes workflow change cheap and continuous, and hire and develop for the taste that decides what all that speed is for.