CIOs want strategic PMOs. I’m not sure they know what they’re asking

In 15 years of PMO consulting, nearly every CIO I’ve worked with has wanted a ‘more strategic’ PMO. And in all that time, very few have been able to describe to me what that would actually look like in practice. Generalities are easy; specifics are hard. They’re even harder when the ground is shifting beneath you.

Every CIO I work with now is caught between two realities. The first: AI will automate most of the coordination, reporting and governance work that has defined the PMO for decades. The second: There is far more to strategy execution and project mobilization than that work ever covered.

Both of these things are true, and together they should be good news. The PMO now has a chance to become what CIOs have always wanted it to be: An engine for change. The problem is that most PMOs have lived in the first reality for so long that they can’t say with any specificity what its next evolution should look like — and CIOs are struggling to articulate it too.

“Be more strategic” is one of those directives that is said often but seldom understood. For most PMOs (and the CIOs that lead them), “strategic” has become associated with big questions rather than specific ones, as if strategy means broad answers to 30,000-foot questions that then get ‘executed’ in the weeds. That’s an awful lot of altitude just sitting there between the vision and the work. And with AI rewriting what the PMO does day to day, the cost of that vagueness is about to go up.

Strategy eventually requires operational specificity.

It requires answering concrete operating model questions about the PMO’s purpose, structure, people, processes, tools and culture, including the fundamental question: “Is a PMO the right way for us to accelerate and govern project work going forward?” These aren’t questions for the PMO director to address alone. CIOs play a critical role in shaping the answers.

Designing for the future: Six questions for your PMO

Together, these six questions form a working diagnostic: If your PMO can answer all six in concrete, specific terms, you have a strategy. If most of them produce vague or aspirational answers, you have a slogan.

Question 1: Purpose

Are we protecting the business cases of our highest-stakes investments — or just tracking their status?

A PMO that exists solely to coordinate and accelerate delivery is already a liability. Solid PMOs today can tell you whether a project is on track. But most aren’t built to tell you whether a project is still a good investment.

Increasingly, forward-thinking PMOs have an expanded mandate. They exist to protect the business case. Protecting a business case answers a harder question: “Given what we know now — about the market, the technology, the competitive landscape, the organization’s capacity — is this still worth doing? And are we managing the risks that could erode its value?”

PMOs that protect business cases do things that reporting-focused PMOs don’t. They flag when the assumptions behind a business case have changed. They surface portfolio-level tradeoffs — what happens to Project B’s timeline and value if we keep funding Project A? They create the conditions for executives to make kill-or-continue decisions before a project becomes too politically expensive to stop. And they give the CIO a fact base for defending those decisions up the chain.

Strategic PMOs protect investment value, not just timelines. If the PMO’s job is to deliver projects on time, it’s an execution function. If its job is to give the organization’s project investments the best possible chance of success — and flag when that’s at risk — it’s a strategic partner. That’s a real choice with real consequences for how the PMO is structured, staffed and evaluated. One requires an army of project delivery managers (or AI equivalents). The other requires a different kind of project leader: One who has the business acumen, analytical skills, judgment, authority and organizational standing to run their project like a business. (And it’s worth noting: Project professionals with high business acumen achieve project business goals 83% of the time compared to 78% for everyone else. They also perform better on budget, schedule and failure avoidance.)

Question 2: Structure

How are we structuring teams to put human/AI capabilities where they make the biggest impact on the portfolio — not just assigning work based on who’s available or what AI is technically capable of?

Most PMOs assign people based on availability rather than on impact. A new project comes in, and teams form around whoever has bandwidth. The question of where each person could create the most value rarely enters the conversation, because the PMO was designed as a logistical, coordinative structure rather than a strategic one.

That’s a problem that gets worse as AI enters the picture. AI agents are already capable of handling much of the coordination, analysis and reporting work that has consumed PM time for years. Many CIOs I talk to think of project management as a binary. Either people are project managers, or agents are.

But the binary framing leads to binary structural decisions: Keep the team as-is or shrink it. Neither version asks whether the roles themselves need to change (or be reinvented completely). The PMOs I see getting this right are putting every role assumption on the table.

“Influence without authority?” That model assumes the PM’s job is to nudge and coordinate. When a PM is accountable for the quality of AI-generated analysis or the integrity of a business case, the question of how much authority they should have gets revisited.

“Temporary assignment?” That made sense when the PM’s job ended at go-live. If the new job has the authority to make delivery decisions with long-term repercussions, it will also need the accountability that comes with a semi-permanent placement (e.g., embedding in a business unit, repositioning as a portfolio manager, among others).

“Only project managers report here?” Strategy execution work is becoming cross-disciplinary. Either PMs will need to become “PMs and something else,” or the PMO will need more diverse roles to support AI-enabled work.

Question 3: People

What capabilities will we need more of, what capabilities will we need less of and what are we doing now to help our people prepare for that shift?

So far, the conversation about upskilling project managers is terribly bland. It centers on improving emotional intelligence, professional judgment and stakeholder management while simultaneously building AI literacy — advice that appears in virtually every PMI publication on the topic and, to be fair, is generally correct. It’s also insufficient.

A more useful version of this conversation looks at the structural decisions above, then asks: “What will each role’s day-to-day look like when AI handles coordination, status collection and routing of work?”

One exercise I use with my clients and in my own practice: I sit down with an LLM (Claude, CoPilot, whatever you prefer) and write a specific prompt: “Here’s everything I believe agentic AI will be able to do by 2028 in a project management context. Given that list, write me the story of what a project manager’s day looks like in that universe. Be specific and descriptive.”

Specificity is crucial. “People skills will matter more” doesn’t help a PMO director build a training plan. “Our PMs need to be able to poke at the implications of vendor pricing models,” or “Our PMs need to understand how to build AI-legible knowledge bases” does.

Once that’s done for the PM role, the next natural step is to run it for every role that touches the portfolio, including any you’re designing from scratch. Then, re-run the exercise every few months as our understanding of AI’s actual capabilities and limitations evolves.

Question 4: Process

With AI handling autonomous workflows, agentic analysis and routine coordination, what are we doing now to prepare our data, artifacts and ways of working for that future?

Most organizations are implementing AI before preparing the work environment AI requires. Conversations about AI in project management today center on what AI can do. Far less of them focus on what organizations need to change about how they work before AI can do it well. Two questions in particular go unasked in the rush to implement agentic solutions, and both have direct consequences for project delivery.

The first is a governance question: How will humans provide meaningful oversight when AI agents are scoping, prioritizing and routing work faster than any team can review those decisions? I’m already seeing this in startup environments. Customer requests are flowing into agentic systems that clarify scope, define requirements, assign priority, generate tasks and route work to human teams. Team A may mark a piece done, but the agent may not catch that Team A needs to talk to Team B about a key decision before Team B’s work begins. The work keeps flowing. The gap compounds.

This creates a tension that most organizations haven’t said out loud yet. The human review layer is exactly the oversight we say we want — and it’s also the bottleneck we’re systematically trying to engineer away. Those two impulses need to be reconciled in process design, not left to sort themselves out in production.

The second is a data question: Are our artifacts, knowledge bases and process definitions in good enough shape for an AI agent to work with? In almost every organization I’ve seen, the answer is no. Project performance data spread across multiple systems, documents and formats. Project requirements in Word documents rather than searchable databases. “1-pager” status reports that were great for executives but useless to AI trying to reconstruct the story of a project. Bloated meeting transcripts used as substitutes for real, contextual information.

These are all processes that “work” today because humans fill the gaps, read between the lines, compensate for inconsistency and apply judgment to ambiguity. AI agents don’t fill gaps the way humans do. And when the data is messy, agents don’t just produce worse output. They get expensive, burning tokens (and budget) trying to make sense of conflicting information, reconciling duplicate artifacts and making choices the data should have made obvious.

The preparation work is specific and unglamorous: Standardizing how project knowledge is captured, structured and maintained so that both humans and AI can effectively search, analyze and act on it.

Question 5: Tools

Do we understand how AI tooling is actually priced, bundled and evolving — well enough to make procurement decisions we won’t regret in eighteen months?

Most PMOs treat tooling as a “process and tools” conversation: What features do we need, what do the demos look like, how does it integrate? That conversation is necessary but no longer sufficient.

AI tooling is introducing a pricing model most PMO directors and IT procurement teams haven’t encountered before. Traditional SaaS is licensed per seat — one user, one license, predictable cost. AI-enabled SaaS increasingly prices two things: Who logs in (the human seat) and what work moves through the system (agent consumption, often metered through credits or usage-based billing).

The specifics vary by vendor, but the structural shift is consistent: The software bill is starting to behave like a hybrid of an access bill and a usage bill — more variable, harder to forecast and tied to throughput rather than simple access to features. (Nate B. Jones has written an excellent breakdown of how major vendors are structuring agent pricing and what to watch for in renewals.)

This shift is already visible in how vendors like ServiceNow, Atlassian and Microsoft are restructuring their enterprise agreements, bundling agent capacity alongside traditional seat licenses in ways procurement teams haven’t seen before.

The implication for PMOs is specific: A team that automates reporting and coordination through an AI-enabled tool may reduce the hours spent on that work, only to then discover that the vendor captured most of that savings through its consumption pricing model.

A PMO that’s being genuinely strategic about tools needs to understand the state of play — how agent pricing works and what to evaluate during procurement. That’s a new competency for most PMOs. Pretending it isn’t will cost many organizations real money.

Question 6: Culture

Are we building a future our people will actually want to work inside, or are we optimizing for efficiency and hoping the human costs sort themselves out later?

Every question above assumes the PMO will have the people it needs to do this work. That assumption is less safe than it used to be.

It’s hard to know how much of the current wave of AI-related layoffs reflects genuine automation and how much is narrative to satisfy shareholders. But the workforce isn’t waiting for the data to come in. Employees are watching what their companies are doing, what they’re asking people to do and what they appear to be getting ready to do. When organizations ask employees to document their own workflows so the company can automate them, the message isn’t subtle — even when the stated intent is to augment rather than replace.

When employees decide the organization isn’t worth investing in, the effects are predictable. Engagement drops. Discretionary effort disappears. Institutional knowledge leaves with every departure, and the people still here stop sharing theirs.

This matters for CIOs specifically because every dimension of PMO transformation described above depends on human judgment. Protecting the business case. De-risking the work.  Evaluating AI decisions. Communicating with customers and stakeholders so they have confidence their needs are well-represented and well-supported. All of this requires a workforce that believes its judgment is valued. Not one that’s bracing for the next round of cuts.

PMOs can become powerful engines for strategy execution. But not inside organizations that are building futures without their people. Workforce shifts are inevitable, and never without casualties. But people are paying attention to who is upskilling, reskilling or providing soft landings for the people affected — and who isn’t.

The decisions CIOs are making right now about how AI and humans work together will shape whether the PMO’s evolution produces an organization people want to contribute to, or one they’re silently planning to leave.

Strategy or slogan?

Every CIO I work with knows their PMO needs to change. The ones getting this right are the ones who hear themselves saying “be more strategic” and push themselves to say what they really mean in specific, operational terms. That’s harder than vision work. But for leaders who like designing things that run well, it’s also the more interesting problem.

None of these six questions has a permanent answer. Technology is advancing, the workforce is shifting and the competitive landscape looks different every quarter. A PMO that answers all six well today will need to revisit them in a year.

But the discipline of asking them — specifically, concretely and without retreating to altitude — enables the people responsible for executing your strategy to stop guessing what “strategic” means and start building toward something resilient, adaptable and useful. And when the next wave of AI capability lands (and it will), you’re not starting the conversation from scratch. Your PMO will be operating against a strategy rather than a slogan.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?