Week 38, 2026
I spent part of a weekend recently reading Product Manager job descriptions, which sounds a lot more exciting than it actually was. One role was in healthcare. Another was in observability. Then cybersecurity. Then developer tools. Then enterprise software. By the fifth or sixth posting, the industries were changing, the company names were changing, and the products certainly weren’t the same. The requirements, though, kept drifting back toward familiar territory.
I wasn’t seeing requests for AI Product Managers nearly as often as the headlines would suggest. What kept showing up instead were things like prototyping with AI, evaluating AI-generated outputs, understanding LLM concepts, working with agents, and using AI tools as part of everyday product work. Two year ago this requirement would have looked unusual. Today it looks normal. That’s probably the story. Not the arrival of a brand new role. The quiet expansion of an existing one.
Back in 2024, most hiring conversations revolved around things every Product Manager would recognize. Customer interviews. Prioritization. Product strategy. Roadmaps. Cross-functional alignment. Launches. Metrics. Business outcomes. The conversation was usually some variation of a simple question: can this person move a product forward? Looking at many of today’s openings, I don’t think that question has changed.
Companies still care about all of those things. They just seem to assume another set of capabilities alongside them. Somewhere over the last few years, a second question entered the room. What have you built with AI? Not what have you read. Not what tools have you tried. Not whether you think AI will transform everything. What have you actually built? That question barely existed a few years ago.
Part of the reason may be that building looks different now. Historically, validating an idea often meant recruiting engineering resources, waiting for capacity, and turning concepts into requirements detailed enough for someone else to bring to life. That’s still true for many initiatives. But the distance between an idea and a proof of concept is much shorter than it used to be.
A PM can sketch a workflow in the morning and have something interactive by the afternoon. Maybe it’s built with Microsoft Copilot. Maybe Cursor. Maybe Claude Code. Maybe one of the many tools that seem to appear every other week lately. The tool matters less than the expectation it creates. Two years ago a strong PRD was often evidence that a PM had done the work. Today a rough but functional prototype can make a stronger case. A prototype is becoming the new proof of product thinking. Not because documents stopped mattering. Because showing became easier.
There’s another change that’s harder to quantify but easier to feel. The vocabulary around Product Management is different. Product discussions that once revolved around adoption, retention, activation, customer needs, and market fit now casually include words like agents, evaluations, hallucinations, RAG, MCP, context engineering, and LLM-as-judge.
I’m not making a case that every PM needs deep expertise in these areas. Most hiring managers aren’t asking that either. But the language itself is revealing. A profession tells you what it values by the words it starts using every day. Nobody talks about cloud literacy the way they did a decade ago because it eventually became part of the background. AI feels increasingly similar. Less like a specialty. More like a capability that’s becoming woven into the fabric of modern product work.
So two years ago, none of these words were Product Manager words, but now they are increasingly differentiators during interviews:
- Autoresearch – allowing agents to run experiments and investigations while you sleep.
- AI agents – use of models to reason, act, and loop through tools.
- AI evaluations – are now part of test suites.
- AI prototypes – now appear alongside specifications instead of after them.
- Context engineering – a topic on its own about everything a model sees before it responds.
- Fine-tuning – now part of product discussion on how to do small tweaks on a base to improve quality and behavior.
- Hallucinations – something PMs now actively measure and mitigate.
- LLM-as-judge – now something you need to think of to evaluate quality at scale.
- MCP – the connection layer you need to manage between models and the tools they need to access.
- Observability – now means that you need to understand why an AI system behaved the way it did.
- RAG – now you need to manage how you are getting context on demand.
- Skills – something that PM now acquires through deep file prompts not repetition.
- Vibe coding- turning English or any other language into working software faster than many of us imagined possible.
What’s interesting is that none of this reduces the importance of traditional Product Management skills. If anything, it puts more pressure on them. AI can help generate options. More concepts. More prototypes. More workflows. More directions to explore. Output is becoming abundant. Judgment remains scarce. That’s still the job. Deciding which customer problem matters. Which signal should be ignored. Which opportunity is real. Which prototype is impressive but ultimately irrelevant. Taste is becoming a competitive advantage.
When I compare Product Manager job descriptions from a few years ago with the ones companies are publishing today, I don’t see evidence that the profession is shrinking. I see evidence that the profession is expanding. The old expectations are still there. Customer understanding. Prioritization. Strategy. Execution. But they’re increasingly accompanied by expectations that would’ve looked out of place not long ago: build with AI, prototype with AI, evaluate AI outputs, understand the strengths and limitations of AI systems, and incorporate them into everyday work. AI is becoming what cloud literacy became years ago. Not a specialty. A baseline capability. And that may be the most important signal hidden in today’s PM job descriptions.
This newsletter reflects my personal views and experiences as a product manager. It does not represent the views, strategies, or opinions of my employer or any organization I am affiliated with.
