Agentic Product Manager: First Mover Advantage
Why I left the programme management playbook behind — and what happens when a PM starts shipping production software with AI.
---
For twenty years I managed programs. Large ones — payments infrastructure at JPMorgan, semiconductor product launches at Qualcomm, enterprise AI adoption at scale. I was good at it. I knew how to decompose ambiguity into workable plans, how to hold a programme together across teams that didn't share priorities, how to deliver complex systems without breaking things downstream.
Then I watched something change that I couldn't ignore.
In early 2026 I started building production software on my own — not prototypes, not demos, but real deployed applications with authentication, databases, and users. The first few sessions were humbling. I was slower than a junior engineer. I made mistakes that would have embarrassed me in a code review. But I was shipping. And with every week, the gap between what I could imagine and what I could execute was closing.
The thing that made this possible wasn't discipline or a new framework. It was agentic AI.
---
What changed
The standard narrative about AI and programming is that it helps developers write code faster. That's true but undersells the shift. What agentic AI actually does — when used well — is compress the distance between product thinking and working software.
For most of my career, the distance between "I understand what this should do" and "this is shipped" ran through a team. You needed a designer, a developer, a QA engineer, a DevOps person. You needed standups and reviews and handoffs. The ideas in my head had to travel through multiple people before they became real, and at every handoff something was lost — context, nuance, the original intent.
With agentic AI, that chain is shorter. I hold the product vision and the domain context. The agent handles execution. The feedback loop — from idea to deployed feature — has collapsed from weeks to hours. This is not a marginal improvement. It is a structural change in what a product person can do alone.
---
The first mover window
There is a window right now — and it is not wide — where people who understand both product and agentic AI have an asymmetric advantage.
Most engineers are excellent at building what they are asked to build. Most product managers are excellent at figuring out what to build. These two capabilities have historically lived in separate people. The person who can hold both — who can think in product outcomes and execute in production code — is genuinely rare right now, and genuinely valuable.
This window will close. In two or three years, the tools will be better, the patterns will be documented, and what feels like an edge today will be table stakes. The people who build intuition now, while the terrain is still unmapped, will have a durable advantage that isn't easily replicated later.
I am not making a bet on a specific tool or model. I am making a bet on a capability: the ability to move from product insight to working system without the usual friction. That capability, developed now, compounds.
---
What agentic engineering actually looks like
It is not magic and it is not easy. It requires a specific kind of discipline that is different from traditional PM work and different from traditional engineering.
The most important skill is not prompting. It is knowing, with precision, what you are trying to build before you start building it. Agentic AI amplifies the quality of your thinking. A vague idea produces a vague system. A well-specified intention — clear scope, clear constraints, clear definition of done — produces something you can actually ship.
The second skill is verification. Agentic systems produce output quickly. That speed is seductive. The discipline is to slow down at the right moments — to read what was built, to test it, to understand it well enough to maintain it. Shipping fast is only valuable if what ships is sound.
The third skill is judgment about when to use the agent and when to think alone. Not every problem benefits from immediate execution. Some problems need to be turned over quietly first — the way a good programme manager sits with ambiguity before naming it, rather than forcing it into a plan prematurely.
---
Why this is a PM problem, not an engineering problem
Engineers who adopt agentic AI become faster engineers. That is real value.
Product managers who adopt it become something different: people who can close the loop between understanding and execution on their own. That is a different kind of value, and in some ways a more disruptive one.
The programme management skills I built over twenty years — stakeholder navigation, risk identification, cross-functional delivery — did not become less relevant. They became more relevant, because now they sit in the same person as the ability to execute. The judgement about what to build and the ability to build it are no longer separated by a handoff.
That is the first mover advantage. Not a technology edge. A capability edge — the result of doing this work seriously, now, while most people are still waiting to see how it plays out.
I am not waiting.