Complex projects fail at a rate most programme leaders already suspect. PMI’s newest research puts a number on it: nearly a third still fail to deliver the full scope of what they promised. The interesting part is not that number, it is what separates the projects that make it from the ones that do not, and it has almost nothing to do with the technology involved.
The Gap Between Success and Failure Is Not Close
PMI’s 2026 Pulse of the Profession report, the sixteenth edition of the survey, found nearly one third of complex projects fail to achieve the full scope of their intended benefits. Eighty one per cent of project professionals say projects have become more complex in recent years, with 37 per cent describing a significant increase. Teams that manage that complexity effectively report an 88 per cent success rate. Teams with minimal effectiveness at managing it report 14 per cent. That gap is the difference between a programme that gets funded again and one that quietly gets shut down.
Why Complex Projects Fail: The Old Reasons Have Not Gone Anywhere
Coverage of the same research and the practitioners interviewed for it lists the specific ways complex projects fall apart, and most of them are not new: unclear governance, competing stakeholder incentives, and decision-making delays. PMI’s Chief of Staff to the CEO, Lenka Pincot, points to sponsorship as the recurring gap. What is really needed, in her words, is active sponsorship, someone who will stand behind the idea and ensure the project stays funded. George Reed, CIO of auntEDNA.ai, points to something even more basic: most projects launch without anyone agreeing what winning actually looks like. It is impossible to succeed if success is undefined, yet executives continue to launch projects without articulating clear, concrete metrics to meet. His suggested test is worth asking on the first day of any programme, not the last: if we were already done, what would winning look like?
The Ownership Problem Nobody Wants to Own
Eric Stettler, a Digital Practice Partner at Kearney, names the failure mode most PMOs recognise on sight: a CIO trying to drive a business decision that was never the CIO’s to make. He calls it a tail wagging the dog scenario, the technology function pushing for change because the business owner never stepped up to own it. Ultimately, as Stettler puts it, the business has to decide whether it is going to operate differently. No governance structure fixes that if the actual business owner has not been named.
The New Failure Mode Sitting on Top of the Old Ones
What has changed is the speed at which a project can now fail. Te Wu, CEO of PMO Advisory, frames the risk plainly: use AI and you get something quickly, and it may look good, but the problem with AI is that it can make things up. A plan, a status report, or a risk register built on AI output that nobody checked line by line does not speed up the old process. It adds a new, largely unreviewed point of failure to a project that already had governance gaps before AI arrived. West Monroe’s Noah Fletcher adds the resourcing side of the same problem: teams are being asked to do more with less, and often more than that with even less than expected, precisely the conditions under which a governance shortcut looks tempting and an unchecked AI output looks like a shortcut worth taking.
What Four Decisions Would Actually Fix
None of this requires a heavier process. It requires four decisions made explicitly rather than assumed: a single named business owner who is not the CIO, an agreed definition of done written down before the work starts, a sponsor who is actively funding and defending the project rather than nominally attached to it, and a checkpoint where a human, not the model, signs off on anything AI produced before it moves downstream, the same accountability gap I examined in a closer look at who actually owns the decisions AI makes inside a transformation programme. Skipping any one of them does not guarantee failure. It just means the project is one of the 31 per cent no matter how the technology performs.
None of the twelve reasons complex projects fail this year would have surprised a programme manager a decade ago, except one. AI has not created a new category of failure, it has just made the old ones faster to reach and easier to miss until the review that should have caught them never happens. The 88 per cent versus 14 per cent gap in PMI’s data is not a technology gap. It rewards the unglamorous governance basics, a named business owner, a sponsor who shows up, a definition of done agreed before the work starts, that most complex projects still skip.
