Drop the armchair strategy: Your AI transformation will fail without the people who know the work

When AI is discussed at the leadership level, the conversation tends to get abstract. People talk about potential, use cases, technological possibility.

But value doesn't appear in strategy documents. It appears in the work itself — across the existing functional boundaries, and in the places where AI opens up possibilities that used to be closed.

The gap between leadership's narrative and the employees' reality is measurable: 76% of leaders believe their people are excited about AI. The actual share is 31% (BCG Henderson Institute / HBR).

This points to a very common and dangerous failure mode in knowledge work. Managers re-designing the work of their employees, without involving them actively. Bain and OpenAI recently named the trap in Harvard Business Review — when companies use AI to optimize existing tasks instead of rethinking the work across functions. They call it the micro-productivity trap — and their diagnosis is right. It's their prescription that is incomplete. Four steps, top-down targets, EBITDA metrics. It assumes something most organizations don't currently have: capacity, courage, and self-confidence to step into the change.

That assumption is what this piece is about.

The train is leaving the station

Today's strategic question isn't whether to get on board with AI. It is how fast you'll be ready to step on the gas. Waiting is not free. At the current pace of development, every week you wait is a week your competitors pull further ahead.

Forget detailed business cases. The business you are trying to model probably won't look the same next quarter as it does today.

The real leadership question is: how do you free up the leaders and employees who are already busy — and give them the resources and support they need to get started without burning them out trying to do their old work while learning the new?

Are you ready to act?

Before you decide how to bring your experience and your position to bear, you have to do two things.

1. Understand your own human biases. When people meet radical change and unpredictability, we get nervous and prefer defensive, safe choices. In this case, thinking primarily defensively is dangerous, because it pulls focus away from the real work: learning, experimenting, and building the new way of working together with employees.
   ‍
2. Investigate what this revolution in knowledge work actually involves. My earlier piece, The AI revolution is a leadership task, is worth reading here, but it won't give you all the answers. AI has such a large reframing potential that you have to see examples before you'll believe it. Set aside hours, not minutes — many people are sharing evidence out there. Find them.
   ‍

Is your organization ready?

Then comes the uncomfortable question: is your organization ready to invest serious amounts of time and mental capacity in adapting itself? How busy are people right now? How high is the stress level? How quickly can you re-prioritize the work and free up the kind of focused time where employees — and you — can learn what the technology can actually do today? (You'll probably be surprised.)

All change follows an S-curve. People have to learn, invest, and adapt before things get better. The question is whether you, as a leader, can persuade yourself, your own leader, and your people to push through the dip.

Unlearn your old mindset

Consultancies and AI vendors can sketch out the right end-state. Most agree that the work has to be redesigned across functions, with the people close to the process. That isn't where it goes wrong.

It goes wrong in the attempt to push that redesign into the classic transformation template: executives set an abstract, bold direction with metrics. Middle managers come back with plans. The steering committee shows up at the milestones. At some point you celebrate that the PowerPoint reports are green, and then you move on to the next transformation.

I might as well say it plainly: that won't work this time.
The AI transformation cannot be communicated, planned, measured, or forecast in the old way. It takes more than that.

Make the AI transformation part of everyday work

As a leader, in the early phase you have to move the AI conversation out of the strategy documents and into the everyday. Four moves make it possible:

1. Frame AI as an inevitable part of the future of work.
2. Frame the technology as harmless in itself, and more accessible than the IT of the past.
3. Create space for ongoing, practical learning and dialogue close to the actual work.
4. Promote experiments that surface both the value of AI for the company and the employee — and where the limits are.

This isn't a project. AI is here to stay. The ability to communicate with machines almost as well as we communicate with humans is becoming a new core competency. That is the democratizing potential of the technology — but it has to be chosen actively.

Once the engine is running and people are experimenting in the everyday, you, as a leader, can step back to the helicopter view — and now you'll see patterns that couldn't be seen from the strategy room.

Because AI doesn't just change the individual work process. It changes the division of labor: who does what, when something gets quality-checked, where decisions get made, how tasks move through the organization. Most often, it also changes how we measure our success. These are structural shifts, and they cannot be designed in isolation from practice.

The consequence is clear: the leader who tries to do all the thinking themselves and roll the redesign out across the few employees that remain is choosing a risky and short-term strategy. What AI cannot bring to your business is judgment, an understanding of the concrete work context, and trust-based connections between departments, companies, and people.

Leadership's job isn't to redesign the work. Leadership's job is to do it together with the employees.

The employees hold the concrete knowledge of how the work actually gets done — where processes break down, where quality comes from, where the hidden dependencies live. If that knowledge isn't brought into play, you'll design solutions that look coherent on paper and fall short in practice.

Leadership still owns direction, prioritization, and framing. But the redesign work itself has to happen in close connection with practice — otherwise it isn't redesign, it is a one-size-fits-all rollout.

Choose the top-down approach and you'll fail. The resistance you'll meet isn't unwillingness — it's fear. It has a name: FOBO, Fear Of Becoming Obsolete. A rational fear of being made redundant. Scared employees find better ways to integrate AI on their own — in secret. Ethan Mollick calls them Secret Cyborgs: employees who use AI every day, but hide it, because the incentives punish openness. The result is a company with no cohesion, no innovation capacity, and little motivation.

That's the blind spot in most transformation guides: they describe the new work from an organizational X-ray, but not from the human emotions that determine whether the redesign becomes reality or a shelf project.

Leadership's job isn't to redesign the work. Leadership's job is to do it together with the employees.

When the work of integrating AI succeeds, something else shifts. AI doesn't just become a tool to be adopted. It becomes an occasion to rethink the work more often, which is what leads to continuous efficiency gains in the long term.

The biggest gains rarely appear where you expect them. They appear when the organization starts working differently — not just faster. I've seen teams at a world-leading company reinvent their campaign development — not because they got a new tool, but because the leadership had the courage to openly name the most important challenges, ask questions instead of giving answers, invite employees in to investigate them through experiments — and let the shared work turn into shared decisions.

The big transformation guides ask: "Which four or five use cases should we pick first?" That isn't a bad question — but it assumes an organization that is already ready to learn. Most aren't, yet. Before you pick use cases, you have to build something more fundamental: a space where employees dare to experiment, share, and fail. That space isn't a project. It is a leadership decision. And you can make it today.

‍