Product in 2026

The cost of turning an idea into working software has collapsed in 18 months. A PM with good judgment and a working AI workflow can prototype, test, and refine ideas at a pace that used to require an engineering pod. I know because I do it. To be clear: I'm not replacing engineers — I'm catching up to them, finally able to put the thing on a screen instead of describing it in a doc.

For twenty years, I've built product by being the connective tissue — translating customer insight into roadmaps, keeping engineering and design and leadership rowing in the same direction, shipping the thing. That work still matters. But the economics of it have changed in a way most PMs are underestimating.

Here's what actually shifted: six months ago, autonomous agents broke after three minutes. Today they run coding tasks for six hours. That's not a feature update. That's a phase change. When an agent can only run for three minutes, your job is to prompt it. When it can run for six hours, your job is to design the system it runs inside. The PM role quietly stopped being about issuing instructions and started being about architecting environments.

Which means: implementation is no longer the rate-limiter. Good ideas are. The new bottleneck is "should we build this?" — a taste problem, a judgment problem, and a customer-understanding problem.

The PMs who win in the next five years are the ones who can:

1. Lead through ambiguity. Set direction when the model's capabilities will shift twice before launch. The PM who waits for clarity before committing has already lost the room.
2. Discover problems before anyone else. Stay pathologically curious about what people do, say, and avoid. Read the forums. Watch the shadow behaviors. Run agents to do the deep research your team can't match — and notice when your own behavior changes, because that's a signal too.
3. Generate quality ideas faster. Quantity is solved — LLMs produce a hundred mediocre ideas in a minute. The scarce resource is the judgment to throw out 97 of them and dream up better ones. Product management is the art of saying no, and AI just made the no-list infinite.
4. Build the MVP. Not because PMs should replace engineers. Because the loop between idea and working artifact is now tight enough to put something in front of real users in a week. Ship experiments. Find winners. Kill losers fast. Only then ask for the engineering investment.
5. Communicate impact crisply enough that partners can recount it. AI can write the doc. It can't make twelve people in a room agree what to do next. Human influence is still the scarce input. Always was.

Here's the part the builder-PM discourse keeps missing. Anyone can swing a hammer. Very few people can produce finished carpentry. AI can put the tools in everyone's hands — Claude Code, n8n, agent frameworks, the whole kit. What it cannot hand you is the taste to know which joint to use, where to hide the seam, when the piece is done. Craft is still rare. Craft is still the job. The tools got cheaper; the skill of producing something worth shipping got more valuable, not less.

My last three years at Google were my education in what these five points look like at industrial scale. Sole PM on a 30-engineer team, I co-built the factuality agent infrastructure with Research — the platform that grounds and evaluates the factuality of generative AI responses across Gemini and Search. We cut ungrounded responses by 50%, brought platform response time from 120 seconds to under 20, and the platform is now used by 20+ teams across Search, Gemini, Workspace, and Research. What I learned building it: the hardest product management problem in AI isn't the model. The model is not the bottleneck. The bottleneck is knowing — with precision — whether the model is doing what you want. Evals are product. Grounding is product. Trust infrastructure is product. Most of the people calling themselves "AI PMs" have never shipped an eval, an agentic system, or a piece of core model infrastructure.

The PMs who will get hired in 2026 are not the ones with the longest resume, and they're not the ones who can rattle off the tool stack. They're the ones who can demonstrate — not claim — that they've extended how they work to meet this moment, and that they still have the product sense to know what's worth building. Tools you can learn in a weekend. Product sense and execution are the work of a career.

The pace of change right now is unlike anything I've seen in twenty years of building product. It's also the most fun I've ever had. If you're building something where AI is the product, I'd love to talk.

— Adam
adam@adamlewkovitz.com