By Sabrina Poteat
May 13, 2026
Every conference, every LinkedIn feed, every software company right now is telling project managers that AI is either going to transform everything or take their jobs. Neither framing is useful to a practitioner who has a real team, a real deadline, or a real deliverable to deliver.
The better question is: where does AI create genuine leverage, and where does human judgment remain the only viable tool?
Over the past four weeks, I had the opportunity to test this directly with 21 participants in their final MBA residency. The project had a defined scope, a hard deadline, and a deliberate decision to introduce AI at specific points while closely observing the results. What follows is a field report based on direct experience.
The Project
Our team was engaged as consultants to help a winery develop a strategy for expanding its business. We were encouraged to use AI throughout the engagement. What we were not given was a blueprint for how. That part we built ourselves, starting with a simple question: given everything this team needed to deliver, where would AI actually make us better, and where would it get in the way?
Where AI Earned Its Place
We used AI in three specific ways, and the approach was the same each time: structured input, defined task, and human review of the output.
The first was project planning. We fed our defined scope into AI and asked it to generate a project plan with dependencies and a timeline anchored to our final presentation date. It produced in minutes what would have taken hours to build manually.
That time did not disappear. Instead of building scaffolding, we were pressure-testing assumptions, identifying risks, and having sharper conversations about the work itself. The scope of work still had to exist before AI could do anything useful, and when dependencies shifted across teams, those negotiations required human judgment about what was truly fixed versus flexible. But AI gave us a head start that changed how we used our time from day one.
The second was team formation. Every member of the team had completed StrengthsFinder, DISC, and a task preference assessment. We fed all that data into AI and asked it to recommend team configurations and task assignments.
It surfaced alignments and gaps that would have taken hours of manual analysis. What could not surface was whether someone was genuinely comfortable with their assigned work or quietly struggling. That required conversation. We had those conversations, and they changed how some work was allocated. AI gave us the structure. Humans made it work for the people involved.
The third was deck development. We uploaded pages of research and gave AI a structure for how we wanted the deliverable organized. It generated a strong first version quickly, which meant we skipped the blank page entirely and went straight to the question that actually matters: is this telling the right story?
The narrative arc, the flow, and the judgment about what the client needed to hear in that room were entirely ours. AI gave us a foundation to react to. We determined what it needed to say.
The pattern across all three is the same. AI absorbed the generative and analytical load, and that freed human bandwidth to move upstream. Less time for construction. More time for decisions about what we were building and why.
What AI Could Not Touch
There is one area that deserves its own space, because no tool, platform, or prompt can reach it.
This team stormed. Conflict emerged, tensions surfaced, and people had to work through it to get to the other side. Twenty-one type-A personalities under pressure do not de-escalate because you ran the right algorithm. And here is the honest version: not all of it resolved cleanly. Some conflict lingered. Some people were better equipped to navigate it than others. Conflict resolution is a skill, and like any skill, it is unevenly distributed across any team you will ever lead.
What mattered was that someone stayed in the room, named what was happening, and kept the work moving forward even when the interpersonal dynamics were not clean. You cannot prompt your way through a room where trust is fractured and timelines are at risk. That requires a human with the awareness to read it and the backbone to address it.
The storming phase is not a problem to be optimized. It is a leadership moment. And leadership is still entirely ours.
What This Means for Project Managers and PMOs
The through line across every use case is this: AI performed when it had clean, structured inputs and a generative or analytical task to execute. It had no role when the work required navigating ambiguity, managing people, or making judgment calls that depend on context a model cannot hold. That is not a limitation to apologize for. It is a design principle to build around.
For project managers, this answers the question surfacing right now. AI can accelerate your planning, process complex team data, and generate starting points that put you in refinement mode instead of construction mode. What it cannot do is read the room, navigate conflict, or make the call that keeps a stakeholder relationship intact. When AI is handling the analytical load, that bandwidth is protected. That is where your value lives, and that is exactly where you should be spending it.
For PMO leaders, the stakes are higher because the decisions are broader. AI can support portfolio planning, resource alignment, and reporting frameworks when the inputs are clean and the methodology is sound. It cannot replace the governance conversations, the prioritization debates, or the moment where someone has to tell a sponsor that their project is not moving forward. But when AI is carrying the analytical weight, those conversations get sharper. You walk in with better information and more capacity to think about what it means.
The leaders who will use AI well are not the ones who adopt most tools. They are the ones who are precise about the boundary, and intentional about what they do with the time and clarity AI creates.
Before layering AI into your project delivery, run it through two questions:
- Is the input clean and structured? If you are feeding ambiguity into a model, you will get confident-sounding ambiguity back.
- Is the output going to a human for review and judgment? AI is a starting point generator, not a decision maker. Remove the human review layer and you have removed the accountability that makes it reliable.
AI is a capable collaborator, not a capable leader. Know the difference and build accordingly.
In the spirit of this article, AI assisted in drafting this piece. The thinking, the experience, and the judgment about what to say and how to say it remains entirely human.
Sabrina Poteat
Sabrina Poteat is the Founder and Principal of Applied Edge Consulting, where she helps organizations design practical, outcome-driven Project Management Offices and improve operational delivery. With more than two decades of experience leading complex programs, building PMOs from the ground up, and driving organizational transformation, she specializes in bringing clarity, structure, and accountability to organizations navigating growth and change.
Throughout her career, Sabrina has partnered with executive leadership teams to align strategy with execution, improve project delivery, and create governance models that support better business decisions. Her work focuses on helping organizations move beyond templates and reporting to build PMOs that provide meaningful insight into priorities, capacity, and execution risk.
Sabrina is the author of Building PMOs That Work and frequently writes and speaks about project leadership, operational excellence, and how organizations can turn project management into a strategic capability rather than an administrative function.
