Perspectives · Leadership & AI
Human Capability Is the Scarce Resource
On October 10 I spoke at the PMI UAE Chapter's annual conference in Dubai, and Martin Bean spoke after me. We had met at breakfast that morning and spent most of it on how close our work had become. Martin is a professor at the University of New South Wales. Before that he ran RMIT and the Open University, co-founded FutureLearn, and wrote Toolkit for Turbulence with the performance psychologist Graham Winter.1 When he took the stage he referred back to my talk several times and named the parallels himself. I want to set them down while they are fresh, because the two bodies of work fit together, and the fit says something about where adaptive capacity gets built and where it gets spent.
My talk covered the tests a project should pass at planning and at execution if sustainability is going to be deliberate rather than accidental, the courage it takes to hold that line as a sponsor, and the energy and water that AI now draws. Martin picked up all three. His own argument started from a sentence I wrote down as he said it: "Enduring human capabilities are the new scarce resource."2
What Martin Bean said about adaptive project leadership
He began by saying that generative AI is unstoppable now, and that the work left to leaders is ethics, values and leadership. As routine cognitive work moves to machines, what will set people apart is judgment, ethics, creativity, empathy, collaboration and learning agility. He cited LinkedIn's finding that 91 percent of learning and development professionals say human skills are increasingly important,3 and LinkedIn's 2026 list of the fastest-growing skills, which is led by cross-functional collaboration, team management, mentorship, public speaking and stakeholder communication.4 He dislikes the term soft skills. They are the hardest skills to master.
The frame under his talk comes from Ronald Heifetz and Marty Linsky.5 A technical challenge is linear. The problem is clear, the solution is known, there is best practice to apply, and fixing it restores equilibrium. An adaptive challenge is the reverse. The problem is unclear, the solution is unknown, there is no best practice, and it crosses every conventional boundary and asks people to change their assumptions and habits. Martin's point for a room full of project managers was that the profession was built on the assumption that nearly every project is technical, and that AI and sustainability are adaptive challenges, both of them.
Then he turned to what people do under that kind of uncertainty. Human beings are wired to go defensive. He listed the behaviors: cautious and fixed, holding tight to control, quick to judge, dismissing other ideas, perfectionism, playing safe, agreeing to anything, avoiding conflict, waiting for others to act, and, the one he called most insidious, passive resistance. Everyone in the room recognized that list. The leader's job is to move people from that posture to one he called constructive, courageous and creative: clear intent, running toward the problem, iterating and expecting mistakes because that is where the learning is.
He closed with six essentials for adaptive project leadership, which I noted as he gave them. Lead with an adaptive mindset. Act with purposeful practices, and pragmatically, because the technology changes every twelve weeks and no course can be set for twelve months. Align sponsors and teams, since his research finds that the gap between the people who sign off and the people who deliver is one of the largest causes of project failure. Build trust and collaborate across boundaries. Coach, don't command. Hold the paradox. The paradox he put in front of the room was AI against sustainability. Efficiency drives more consumption. AI learns from yesterday's data while sustainability needs a future unlike the past. The model recommends in seconds while the impact lasts for generations. And AI automates the entry-level work, the first rungs on the career ladder, so the profession's own next generation is at risk. He did not cite a figure for that last point, and there is one. Stanford's Digital Economy Lab reported in August, from ADP payroll records, that workers aged 22 to 25 in the most AI-exposed occupations are 19 percent below the trend of their less-exposed peers, mostly because fewer are being hired.6
Where his model and mine meet
My own work on this is the Human Adaptive Capacity Model for AI, which I set out in these pages in September. It adapts the foundation-and-ceiling logic of Kate Raworth's Doughnut Economics to technology. The inner boundary is the human capability foundation, the minimum people need to take part in an AI-shaped world and keep their own judgment while doing it. The outer boundary is the human adaptive ceiling, the rate of technological change above which people and institutions lose competence, oversight and recovery capacity faster than they can rebuild them. Between the two is the regenerative operating space, where AI leaves people more capable than it found them. Two expressions run the model. The adaptive pressure ratio is technological pressure divided by adaptive capacity. The adaptation deficit is the rate of AI capability growth minus the rate at which people adapt.
Martin's enduring human capabilities are my foundation ring in different words. Judgment, ethics, creativity, empathy, collaboration and learning agility sit on three of its eleven dimensions: professional competence and practice, education and lifelong learning, and human agency and judgment. His scarce resource is my claim that adaptive capacity is finite. His twelve-week clock set against a twelve-month planning cycle is the adaptation deficit in a sentence a project manager can act on. His four generations in one workplace are a reminder that the foundation is not met evenly across a population, which is the model's equity and fair access dimension. His point about the first rungs of the ladder is the argument I made in September about the supply of senior judgment. Each firm that stops hiring juniors makes a decision that is rational for the firm, and nobody is accountable for the senior professionals who will not exist in ten years.
The difference between us is scale, and the difference is the useful part. Martin works on the person and the team: the mindset a leader walks in with, the behaviors to watch for, the operating rhythm, the coaching conversation. I work on the rate: whether the pressure on people is growing faster than their capacity to absorb it, and how an organization would measure that. His model is what a leader does on a Monday. Mine is how the organization finds out, a year later, whether Monday was enough.
His model is what a leader does on a Monday. Mine is how the organization finds out, a year later, whether Monday was enough.
Coaching is how adaptive capacity is rebuilt
Of his six essentials, the one I would put first is the one he put last: coach, don't command. In September I wrote that competence comes from one loop. You do the work, you get it wrong, someone corrects you, you do it again. When AI does the reasoning and a person reviews the answer, that loop runs less often than the job used to make it run. Coaching is the loop run on purpose. A coach puts the junior in front of the problem before the tool, lets them get it wrong where it is safe to be wrong, and corrects. Martin's phrase for it is being the coach your people need. My phrase for the result is regeneration: the people who use a system are more capable afterward, measured with the system switched off. The two phrases describe the same practice from two sides.
Command does the opposite, and so does unmanaged AI adoption, for the same reason. Both take the decision away from the person who needs the practice. Martin said command and control was probably never a great leadership strategy and will not work at all with the generations now entering the workforce. I would add that it also drains the foundation. Every decision taken away from a junior is one fewer repetition of the loop, and the deficit compounds.
Coaching is the loop run on purpose. Every decision taken away from a junior is one fewer repetition of it, and the deficit compounds.
What the two models give a sponsor
Put them together and a sponsor has a test at each end of a project. At the front end, Martin's test: whether the challenge is technical or adaptive, and if it is adaptive, what the leader will do about the defensive behaviors that are coming. At the back end, mine: whether, twelve months on, the people on the project can do and check the work the system did, with the system off, and whether more of them can than at the start. The first changes how the project is led. The second tells the organization whether the leading worked. Neither appears in a standard business case for AI, and the PMI® GPM® Guide to Responsible Project Sponsorship puts the authority to ask for both with the person who signs.8
What struck me at breakfast was that two fields had arrived at the same limit by different routes. Leadership development got there by watching people under pressure. Sustainability got there by measuring systems asked for more than they could give. The limit is the same. Human capability is the resource being drawn down, and the business case for AI does not yet have a line for it.
About the author
Dr. Joel Carboni is the founder and president of GPM and the architect of the standards, methods, and certifications behind Sustainable Project Management®, delivered through the PMI-GPM joint venture. He created the P5 Standard, the Sustainability Competence Standard, the Sustainable Project Management® Practice Guide, the Project Sustainability Reporting Guide and the PMI GPM Guide to Responsible Project Sponsorship. For over 30 years, his work has moved the profession in a different direction. A Forbes Business Council member and Global Reporting Initiative (GRI) contributor, he was shortlisted for Thinkers50's inaugural regenerative business award in 2025 for his book Becoming Regenerative.
1 Winter, G., and Bean, M. (2023). Toolkit for Turbulence: The Mindset and Methods That Leaders Need to Turn Adversity to Advantage. Wiley. thebeancentre.com
2 Bean, M. Keynote and panel introduction, PMI UAE Chapter annual conference, Dubai, October 10, 2026. Quotations are from the author's notes and the session recording.
3 LinkedIn Learning, 2024 Workplace Learning Report: 91 percent of L&D professionals agree that human skills or soft skills are increasingly important. linkedin.com
4 LinkedIn, "Skills on the Rise: The Fastest-Growing Skills in 2026," February 24, 2026. news.linkedin.com
5 Heifetz, R. A., and Linsky, M. (2002). Leadership on the Line: Staying Alive Through the Dangers of Leading. Harvard Business School Press. The distinction between technical and adaptive challenges first appears in Heifetz, R. A. (1994), Leadership Without Easy Answers, Harvard University Press.
6 Stanford Digital Economy Lab, August 2026, analysis of ADP payroll records covering workers aged 22 to 25 in AI-exposed occupations.
7 Carboni, J. B. (2026). "The AI Slowdown Debate Leaves Out Human Adaptive Capacity." GPM Perspectives, September 2026. The model: Carboni, J. B. (2026). The Human Adaptive Capacity Model for AI. Conceptual framework; 11 foundation dimensions, 13 ceiling pressures and 8 overshoot consequences, each with an illustrative indicator. Thresholds are not yet calibrated. gpm.org/perspectives
8 PMI® GPM® Guide to Responsible Project Sponsorship (2026), a PMI-GPM publication. gpm.org/sponsor