AI strategy and AI governance

AI adoption is behaviour change

The biggest barriers to AI are human: how people work, what they trust and what they worry about. That is also where the most can be gained.

· · 3 min read

Novo Nordisk rolled out generative AI to 20,000 employees, one of the largest roll-outs so far. The hardest problems turned out to be about behaviour, trust and leadership. I recognise that from practice.

Five assumptions that don't hold

IMD researchers and the people responsible at Novo Nordisk describe five assumptions that slow adoption down.

"It's all about efficiency." Employees saved an average of 2.17 hours a week. Yet their satisfaction was three times more strongly linked to the quality of their work. For most people, better work counted for more than faster work.

"Adoption grows by itself." After about three months, usage dropped. Those who kept going achieved more later on. Well-timed training, real use cases and champions brought the momentum back.

"Younger staff will lead the way." It went the other way: experienced employees gained more in productivity and quality. They see more quickly where AI helps and are better at judging whether an output is right.

"A digital culture has no resistance." Some people worried that colleagues would see them as lazy or unethical for using AI. Talking about it openly, room to practise and clear guidelines help.

"One roll-out fits everyone." A finance team needs different things from a sales team. Tailoring by function worked better.

Don't promise that AI won't cost jobs

Amit Joshi warns leaders against a well-meant reflex: promising that AI won't cost jobs. Employees don't believe it, and that distrust slows adoption. He argues for an honest conversation about how roles will change, for explaining how AI reaches an output and for a concrete answer to what it brings each role. People also need a safe space to experiment.

Joshi also points to "shadow AI": tools people use on their own initiative, often on their personal phones. Ignoring it solves nothing. You are better off bringing it into the open and agreeing how it is used.

Four questions for responsible AI

Trust also needs a moral compass. Hoque, Sanklecha and Scade propose four questions to keep asking at board, programme and technical level:

  1. What does it bring us, including in the long term?
  2. What are the likely consequences for everyone involved?
  3. Which principles does this support or violate?
  4. What does this say about who we are?

Their example is IKEA, which reskilled employees whose work was changed by AI rather than letting them go. A choice like that shows what kind of organisation you are.

What I advise leaders

Start small. A few simple applications that people use every day build trust faster than one big project. Find colleagues who like to lead the way and give them room to share what works and answer questions as champions. A network of champions is the strongest driver of adoption I know. Make good examples visible, such as shared prompts, so others don't have to reinvent the wheel.

And record a baseline before you start. Without one you cannot show later what it delivered, and scaling up becomes a debate about gut feeling.

Sources

  1. Wade, M. R., Trantopoulos, K., Navas, M. & Romare, A. (2025). 5 myths holding back GenAI in the workplace. I by IMD, 19 August 2025.
  2. Joshi, A. M. (2026). Bosses: Stop telling staff that AI won't take their jobs. I by IMD, 8 April 2026.
  3. Hoque, F., Sanklecha, P. & Scade, P. (2025). Four questions to guide the implementation of responsible AI. I by IMD, 25 July 2025.

Read next

Let's talk

Would you like to exchange ideas on data, analytics, AI or FP&A? I would be glad to connect.