Every PMO leader I meet is keeping a private score on how far behind they are on AI. But the yardsticks most of us use to measure ourselves are broken.
Every workshop, keynote, and client engagement I’ve run over the last two years has ended up in the same place. We don’t get to talk about running a PMO anymore (not responsibly, anyway) without talking about AI.
It doesn’t seem to matter who’s in the room, either. The leader running agentic teams every day feels behind. So does the one still working out how to write a decent prompt.
For my part, I’m what the kids call “agent-pilled”. I’m building an agent twin that works the way I do, so she can take first passes at my tasks before they get to me. Her name is Belinda. That’s what I do on weekends. And I still feel behind.
Everyone feels it.
Every PMO leader I talk to circles the same thing. Is my PMO behind on AI? I’ve come to believe that’s almost never the question they came to ask. It’s standing in for a smaller, more uncomfortable one.
Am I behind too?
Is your PMO behind, or are you?
When someone asks whether their PMO is behind, I’ve started asking what they’re picturing. The answers are consistent. Usually some mix of these.
- Questions about relevance. What can AI do today, what will it do next, and what does either answer mean for my PMO’s relevance and for the careers of the PMs I lead? If I teach my PMs to use AI well, am I making them stronger or making them replaceable?
- Questions about credibility. Will we be ready for the day IT finally opens the gates and lets us use real tools? Can my PMs use a word like “agent” correctly in front of the people who are sizing up whether I know what I’m talking about?
“We all want to know. Am I relevant? Am I credible?”
Those are questions about a PMO. But they’re also about the person asking them. We all want to know. Am I relevant? Am I credible?
What makes leading a PMO hard right now is that the organizational questions come due at exactly the moment we’re worrying about the personal ones.
AI isn’t the internet, and it isn’t the steam engine.
We’ve been here before, the skeptics say. Every technology revolution shows up alongside a prediction that some category of people is about to become unnecessary. The steam engine. The PC. The internet. And the prediction is usually loud, usually specific, and usually wrong in the details.
And they’ve got a point. Big social shifts usually arrive more slowly than we predict, and the jobs we were sure would vanish tend to change shape instead of disappearing. David Autor and Neil Thompson’s research across hundreds of occupations found automation cuts more than one way. Sometimes it removes the need for people. Sometimes it concentrates a role around the tasks hardest to replace, and the expertise left standing can be worth more. For better and for worse. The mix should keep us calm.
But I don’t think AI is like the internet, or like the steam engine. Those were tools. This is a new kind of intelligence. It isn’t human. And it isn’t just a faster or better-connected version of something we already had.
“We can’t predict where this goes. We can predict the pace.”
I’m not saying that to sound ominous, and I’m not selling optimism either. My claim is narrower than either of those, and I think it’s harder to argue with. We can’t predict where this goes. We can predict the pace.
By METR’s 2025 measurements, the length of software task a frontier AI agent can finish on its own (at 50% reliability) doubled roughly every seven months for six years.
It’s fast.
We don’t need to predict AI to know what to do about it.
Here’s the part I think is good news. We don’t have to know how this technology will evolve. (Which is lucky, because prediction is a losing battle.) Whatever you believe about where AI is going, your next step as a PMO leader is the same.
Learning to use the internet in the 1990s was never a bet on the internet. It was just what you had to do to keep working. Agile had a slow start too. Today roughly 60% of the PMP exam is built around agile and hybrid approaches, and, in my experience, a serious PM role now expects fluency in both. Nobody who picked up Agile in 2015 had to forecast any of that correctly. They just had to not sit it out.
We’re comparing your PMO to the wrong companies.
Compare your PMO to a small tech company and you’ll always come up short. You’re also being too hard on yourself. Those companies have almost no constraints. Nobody is asking them to scale a solution across forty business units, or collect a thousand approvals just to download and try something new.
And they’re often staffed wall to wall with developers who can push AI to its limits in ways most of us can’t. Watching them is worth your time, but only as a thought experiment. If this is what’s possible with the brakes off, what might a PMO look like in three years?
As a consultant, I watch those companies closely and copy what works (hence, “Belinda”) so I can help answer exactly that question. What do my clients need to know, do, decide so they are ready when this becomes their reality? I’m not telling PMO leaders to go build Belindas. That’s silly. There’s a portfolio to run.
Compare yourself to a big, heavily regulated enterprise and the comparison flips. All that bureaucracy, all those security reviews and scaling considerations, and suddenly you feel right on time.
That’s the more dangerous comparison, because the comfort is false. If that’s your world, you owe it to yourself to read a little about what’s happening beyond the walls of the Copilot chat castle. When the gates open (and they will), a PMO whose entire experience of AI is the one sanctioned tool won’t be ready.
(CIOs, if you’re listening. This is your strategic problem to manage. Keep your orgs safe. Move at a responsible pace. But understand that each quarter your PMO is running on rudimentary AI today will cost you triple that playing catch-up later).
Each comparison is good for exactly one thing. The small company shows you the ceiling, what becomes possible when nothing is in the way. The big one is a warning against letting your own constraints talk you into looking away. Neither one is your benchmark. You need a different measuring stick.
Stop obsessing over the AI ceiling. Keep up with the floor.
The ceiling moves every week. Chase it and you end up exhausted and permanently convinced you’re behind. You don’t need to sit at the frontier of AI, especially not in what you use every day.
Keep up with an intelligent, strategic floor instead. Here’s mine. You’re on solid ground if most of these get a yes.
Inside Your Org:
- You’re ahead of most non-technical people in your organization. If they’re prompting, you’re building a library of repeatable prompts. If they’ve got saved prompts, you’re building skills on top of them. Being a step ahead is what puts you in a position to help them.
- You’re ahead of most people on your own team. (Unless you have one true nerd. There’s always one. Use the nerd to your advantage.) “Ahead” here means you’re putting the available technology to work for you, either because it’s valuable on its own or because learning it is.
Your PMO gets evaluated on its relevance and usefulness to your organization, not on its general AI knowledge or competence. To be of real help, you need to know just a little bit more than the other non-technical people around you.
Inside Yourself:
- You read and understand major headlines from OpenAI, Anthropic, or Microsoft. Bonus points for other platforms or threads, though those are the big three labs most enterprise organizations are adopting in one form or another. Start with them.
- You’re curious enough that you haven’t used AI the same way for twelve weeks running. If your usage today looks like your usage in the spring, you’ve stopped exploring. Write a new prompt, build a new skill, set up a project space a different way. Small experiments, happening often.
If reading that list makes your chest tighten, you’re not alone. But this is the job now, the way reading a schedule or writing a report always was. Executives already rate project management lower than we do. A PMO leader who can’t hold a credible conversation about AI loses executive trust on top of that. That will only get more true.
If it helps, my read is that claiming a few of these puts you in the top 30% of PMO leaders I speak with. The “floor” I’ve outlined here is ambitious, but it works. Clear it and you’re easily in the top 10%, which is where you want to be if your leaders are counting on you to lead your PMO assertively into the future.
You can’t read your way onto the frontier.
There’s one real difference between building AI competency and everything else this profession has asked you to learn. The rest of it arrived with a canon. You picked up scheduling, reporting, what Agile is, by studying what other people had already worked out, written down, and built exams around.
“Nobody has the manual, because nobody has written it yet.”
AI has almost none of that yet. It’s exploratory and ambiguous. That calls for a different kind of thinking, and more time to absorb. The strange comfort is that everyone is figuring it out at once. Nobody has the manual, because nobody has written it yet.
It takes some bravery, because right now the only real way to learn is by doing. That’s wildly effective, and it can also be inefficient. You’ll mess up, start over. Have an idea, build something, realize it sucks, scrap it. Build something great, then a new product renders what you built irrelevant.
I bring that up to take the sting out of it.
Now the uncomfortable part, for the reader whose security team says no to everything, and whose only sanctioned tool is whatever the company already picked. (Who gets to decide that is a whole separate question.) Some of this you’ll have to do on your own account, on your own time (and on non-confidential tasks).
I know how that reads. But think about how we already treat our own growth. We pay for our own certifications when we think they’ll improve our odds. We buy the book, we take the course. This is that. There’s a lot you can do on a $20-a-month subscription. More if you’re paying for two or three (I lean on Claude and Perplexity). More still on a $100 or $200 plan, where building agentic teams comes within reach.
Point them at your personal life rather than work. The skills are identical, and unlike a certificate on the wall, you get real value out of it in the rest of your life while you’re learning.
If You Only Do One Thing
Ahead or behind on the floor criteria above, the starting place is the same.
Use AI a little differently each coming month. Build experimentation and incremental learning into how you work. Don’t let yourself look up in April and realize you’re using AI exactly the way you did in January.
Build the habit of learning by doing and the other competencies follow on their own. And I think you’ll find it’s pretty fun.