September 24, 2026

What is adaptive leadership, and why does AI make it urgent?

by Colin Cooper

SHARE THE ARTICLE

The short version

Adaptive leadership is for problems with no known fix, where the people with the problem have to change for it to be solved.
Technical problems have answers an expert can supply. Adaptive challenges need learning, so the leader gives the work back to the people who own it.
The practice runs as a loop: observe, interpret, intervene, then observe again.
I’d argue AI is turning more work into adaptive challenges. Employers in a World Economic Forum survey expect 39% of workers’ skill sets to change or go out of date by 2030 (World Economic Forum, 2025).
I count Adaptive Leadership as one of five human intelligences AI can’t automate.
The model has fair critics, mainly that it’s been described far more often than it’s been tested.

This guide covers where the model comes from, its steps and principles, and why AI raises the stakes. After that, you’ll find how I practise it, the criticisms and a first step for this week.

Where adaptive leadership comes from

Adaptive leadership comes from Ronald Heifetz at Harvard Kennedy School, who developed it with Riley Sinder and Marty Linsky. He first set it out in Leadership Without Easy Answers in 1994, a book his school calls “a classic in the field” (Harvard Kennedy School, 2026).

In 2009 Heifetz, Linsky and Alexander Grashow followed it with The Practice of Adaptive Leadership, a hands-on, practical guide from Harvard Business Press (Harvard Kennedy School, 2009). The same year they wrote in Harvard Business Review that “adaptive leadership is an improvisational and experimental art” (Harvard Business Review, 2009).

That word, art, is one I’d hold on to. Adaptive leadership’s something you practise and get better at. It’s their model, so I’ll set it out as they describe it first, and then show you how I use it.

Technical problems and adaptive challenges

A technical problem has a known answer that an expert can supply. An adaptive challenge doesn’t, so the people who have the problem have to learn their way to an answer. A Harvard Medical School guide to the model describes the adaptive kind as “tied to deeper patterns or dynamics”, one that “requires learning” (Harvard Medical School, 2021).

Heifetz and Donald Laurie used a doctor and patient to show the difference. The doctor could supply the technical expertise. Only the patient could change the habits that were hurting his health (Harvard Business Review, 1997).

Now put AI in the room. Say you run a 40-person firm. Connecting an AI tool to your CRM is technical, and you’d get your IT partner to do it by Thursday. Deciding which client calls stay human, who checks the machine’s work, and what your juniors learn now that first drafts write themselves is adaptive. Nobody can hand you that answer, because it’s in your people’s habits and judgement.

It’s a costly mistake to treat the second kind like the first. You buy another tool and book a training day, then wonder why nothing’s changed.

The three steps: observe, interpret, intervene

The three steps are observe, interpret and intervene, and you run them as a loop rather than a checklist. Harvard Medical School’s guide sums up the practice as “Observe, pause for reflection, interpret, then intervene” (Harvard Medical School, 2021).

Observing means getting on what Heifetz and Laurie call the balcony. You’re stepping back far enough to spot the patterns, rather than only seeing the next task (Heifetz and Laurie, 1997).

Interpreting means making a guess about what’s really going on, and holding it loosely. Intervening means trying something small. Then you’re back up on the balcony to see what it did.

It’s tempting to jump straight to step three, because fixing feels like leading. Don’t. Skip the first two steps and you can end up fixing the wrong problem very efficiently.

The six principles behind the model

Heifetz and Laurie set out six principles in Harvard Business Review in 1997. Get on the balcony, identify the adaptive challenge, regulate distress, maintain disciplined attention, give the work back to people, and protect voices of leadership from below (Heifetz and Laurie, 1997).

I’d say the last four are the harder ones, and they’re where most of the work is. In my words, regulating distress means keeping the heat on without boiling anyone. Disciplined attention means you don’t let the team slide back into the easier, technical conversation. Giving the work back means the team solves it, not you. And protecting voices from below means the junior who spots the real problem doesn’t get shut down for saying so.

You’ll also see four principles listed online: emotional intelligence, organisational justice, development and character. It’s a fair summary of the attitude. But it isn’t the list in Heifetz and Laurie’s article, so when you’re applying the model, work from the six.

Why AI makes adaptive leadership urgent

As I see it, AI’s taking on the work with known answers. What it leaves behind is the change only your people can make.

The World Economic Forum’s Future of Jobs Report 2025 surveyed more than 1,000 employers across 55 economies. They expect two-fifths (39%) of workers’ existing skill sets to change or go out of date between 2025 and 2030 (World Economic Forum, 2025). Analytical thinking is still the core skill employers want most. Resilience, flexibility and agility are expected to keep rising in importance, along with curiosity and lifelong learning.

For a leader, that’s a list of adaptive challenges. A new tool is a technical problem. But a team whose skills are shifting under its feet is an adaptive one. In 2009 Heifetz, Grashow and Linsky were writing about a different crisis. They argued that leaders should embrace disequilibrium and use turbulence to build new capacity (Harvard Business Review, 2009).

Your tools will change every six months, and adaptability is the only skill with a long shelf life. That’s my view from the work rather than from a study. But it’s why I’d put more into how your team learns than into which tool it learns. It’s also why I talk about putting AI in its place: human first, machine second.

Adaptive Leadership is a human intelligence AI can’t automate

I count Adaptive Leadership as one of the five human intelligences AI can’t automate, with Curiosity, Creativity, Critical Thinking and Emotional Intelligence. Together they’re the 5 Pillars of Human Intelligence. They’re my framework, built from my work rather than from a lab, so treat them as an approach.

I’ve hired and trained more than 1,000 people, across distributed teams built since the early 2000s. When I hire, these five are what I’m really testing for.

AI can draft a change plan for you. It can’t sit with a team while that team lets go of the way it’s always worked, or decide what it should care about next.

In my own work, Adaptive Leadership’s home is the Pivot Formula, which asks one question: what must I learn, release or reinvent? That question does the same job as the balcony. It makes you look at what’s got to change in people as well as in the plan. That’s why I keep coming back to one line: “AI masters what already exists. But humans are the architects of what’s next.”

Leading without force or authority

Leadership Without Force is my approach to leading through calm authority rather than force. It grew out of around 15 years of natural-horsemanship clinics in Australia.

You can’t force a horse, and you can’t force a human. Either you earn trust or you earn nothing. That’s the heart of it, and it sits neatly beside Heifetz and Laurie’s principle of giving the work back to people (1997). The horse has to make the change itself, and so does your team.

This matters more with AI, because some of the people you’ll need to bring with you won’t report to you. They’ll be suppliers, or colleagues in another team. Titles don’t create leaders. Behaviour does.

It sits alongside my authentic leadership work, because calm authority only works when people trust who you are. I’m writing a separate guide on leading without authority, with more on what the horses taught me.

Make it safe to learn

Adaptive work asks people to try things that won’t always work. Your job is to keep the pressure high enough to move them and low enough that they’re still thinking clearly. Heifetz and Laurie call this regulating “the inevitable distress that adaptive work generates” (1997).

Name the change out loud, so nobody’s left guessing. Break the work into small experiments, so a failure costs a week rather than a year. And when one does fail, use the four-step personal pivot from my Pivot Formula talk. Pause, to get some distance from the reaction. Interpret, by separating facts from emotion. Validate, by taking the learning without the blame, and that one takes practice. Then Optimise, by changing the next action.

I’ll cover psychological safety when AI is changing every job in its own guide. There’s also a piece on why fear of failure stops people learning.

Run adaptive work in 12-week cycles

I run adaptive work in 12-week cycles. An annual plan assumes you already know the answer, and an adaptive challenge tells you that you don’t. Twelve weeks is long enough to try something real and short enough to change course before it gets expensive.

The Pivot Formula runs each cycle through five levels: IDENTIFY (clarity before strategy), DEPLOY (execute with agility), REFLECT (measure and learn), SCALE (expand without breaking) and AUTHORITY (lead the market). IDENTIFY and REFLECT are where the balcony work happens, and if you skip them you’re back to fixing.

For your own diary, the 12-Week Success Clock is the personal version, a productivity operating system for one 12-week cycle at a time. The idea fits in one line, and it’s the one I close my 12-week keynote with: “Stop planning in years. Start finishing in weeks.”

There’s more on planning a year in 12-week cycles in a separate guide.

What it looks like in schools, universities and L&D teams

In a school or university, I’d treat AI as an adaptive challenge before a technical one. Buying a licence is the quick bit. Agreeing what good learning and fair assessment look like now is work only staff and students can do together.

So the leaders I’d back are the ones who give that work to their teachers and protect the time to do it. I made the education case in my TEDx talk on AI and VR in schools. I’ve also looked at whether AI will replace teachers in its own piece.

For L&D teams, the split’s the same. Training people on a tool is technical. Building a team that can teach itself the next tool is adaptive, and that’s the part worth your budget.

The criticisms worth taking seriously

The fairest criticism is that adaptive leadership’s been described far more often than it’s been tested.

In 2023, researchers who built a questionnaire to measure adaptive leadership behaviour wrote that most earlier work on it was “only theoretical and remains rather abstract”. They said the idea still needed “empirical scrutiny” (Nöthel and colleagues, Frontiers in Psychology, 2023).

A 2024 meta-analysis brought together 31 studies with 11,640 participants. Leadership and employees’ adaptive performance were moderately linked, but “no differences emerged from the different leadership styles” the studies looked at (Bonini and colleagues, PLoS One, 2024). So good leadership and people’s ability to adapt tend to go together. The evidence doesn’t yet show that one named style does it better than another, mine included.

There’s a personal cost too. Harvard Kennedy School’s description of that book says that challenging people’s familiar reality “can be difficult, dangerous work” (Harvard Kennedy School, 2009), so expect to take some heat.

And not everything’s adaptive. If the printer’s broken, fix the printer. Save the balcony for the problems that need it.

Your first step this week

Take ten problems sitting on your desk, mark each one technical or adaptive, and pick one adaptive challenge to run as a 12-week experiment. If you’d like to build it into a fuller leadership development plan, I’m covering that separately. For now, four steps:

Sort them (20 minutes). Write ten problems down and mark each one T or A. If an expert could fix it by Friday, it’s a T.
Pick one A. Write down who has to change, and what they’d have to give up. That second line’s often the real problem.
Book the balcony. Put 30 minutes in your diary every Friday for the next 12 weeks. Use it to observe and interpret, and don’t fix anything yet.
Start with questions. For the first fortnight, ask and listen, and hold back on answers. Then try one small change and watch what it does.

How Colin helps

I’ve spent nearly three decades building and running businesses. Today I help leaders, founders and educators build what AI can’t replace, through keynotes, advisory work and programmes. For an adaptive challenge, that’s usually advisory work through the Pivot Formula. Or it’s a keynote that gets a leadership team or staff room facing the same problem.

If you’d like to talk about a keynote or advisory work for your school, university or business, get in touch.