Ask a project professional whether projects feel harder than they used to, and you’ll get a knowing look. PMI’s 2026 Pulse of the Profession report puts a number on it: 81% of project professionals say projects have become more complex in recent years. We doubt many readers are surprised. The more useful question is what to do about it.
Complexity isn’t just more work
More tasks is a scheduling problem. Complexity is different: many parts that affect each other in ways nobody planned. A change to one system alters a process in another department, which changes how customers behave, which changes demand on the first system. Plans built as lists of tasks handle this badly, because the trouble sits between the tasks.
That’s why we keep arguing for systems thinking. In practice, it means three habits:
- Map dependencies, including the informal ones. Who needs to know when this changes?
- Look for feedback loops. What happens when this success makes demand rise, or this fix creates a new queue?
- Ask about second-order effects. If this works, what does it change next?
Run these questions at the start of a project and again at each major decision. They take an hour, not a week.
The AI disconnect
The same report notes a gap worth reading twice. Leaders frame AI as an external disruption, something happening to the market that the organization must respond to. Project teams experience it as internal execution strain: new tools to learn, new outputs to check, and new expectations on top of the same deadlines.
Both views are accurate, and the gap between them is where trouble starts. A strategy slide that says “adopt AI” becomes a project team’s extra workload, with no added time to learn the tools or review what they produce.
Treat AI adoption as a delivery change
We’d suggest leaders handle AI the way they’d handle any significant change to how work gets done. Picture a mid-sized logistics firm that rolls out an AI assistant for planning and reporting. Three things need to be in place.
Capacity means time in people’s schedules to learn, experiment, and fix mistakes, not just a license. Training means practical sessions built on the team’s own work, not a general overview. Governance means clear rules on what can be automated, who checks the output, and where data may go.
Without those, you’ve announced a strategy but not delivered a change.
Where to start
Pick one AI initiative currently underway. Sit down with the people doing the work, not only the sponsor, and ask what has become harder since it started. Then draw the dependencies on a whiteboard. You’ll almost certainly find something the plan missed, and that’s the point.