Executive AI role-modelling is the practice of senior leaders using AI visibly in their own work, so that the organisation sees what is permitted, what good use looks like and that it is safe to learn in public. It is an adoption lever: staff take their cue from what leaders do with AI far more than from the strategy deck.
The idea goes by several names: leading by example with AI, visible leadership use, leader-led adoption, “prompt first” leadership and, in the phrase used in the conversation this article draws on, using AI out loud. Each describes the same mechanism: trust is built by watching respected people do it, including where it goes wrong.
This is not a soft issue. ISO/IEC 42001:2023 requires top management to demonstrate leadership and commitment to the AI management system, not simply to approve it. The NIST AI Risk Management Framework (AI RMF 1.0) treats organisational culture as part of its Govern function, expecting teams to be committed to a culture that considers and communicates AI risk. The EU AI Act (Regulation (EU) 2024/1689), in Article 4, requires providers and deployers of AI systems to take measures to ensure a sufficient level of AI literacy among their staff.
The evidence on leadership behaviour is striking. The share of employees who feel positive about generative AI rises from 15% to 55% when they have strong leadership support (Source: BCG, AI at Work 2025: Momentum Builds, but Gaps Remain, 2025). Yet just 25% of frontline workers say their leaders provide enough guidance on AI (Source: BCG, AI at Work 2025: Momentum Builds, but Gaps Remain, 2025).
This analysis builds on a recent conversation with Matt Berg, Director of AI Business Solutions, Business Strategy at Microsoft, about how everyday leadership habits shape AI culture.
Why does AI adoption depend on what leaders visibly do?
AI adoption depends on visible leadership behaviour because employees are uncertain about what is allowed, and they resolve that uncertainty by watching senior people. A licence tells staff that a tool exists. It does not tell them whether admitting AI helped is a strength or a confession, or whether a mistake will be forgiven.
Berg’s first piece of advice to leaders is direct: use AI out loud. In practice that means telling a meeting that AI reread the last call’s notes and surfaced something that had been missed. The signal is that AI use is normal, expected at the top, and that the leader still owns the judgement.
Berg also makes a point executives should find reassuring: “The reality is you can’t know everything. But there’s nothing wrong with that.” Leaders who admit they are still learning give their teams permission to do likewise. Leaders who stay silent, or who delegate AI entirely to IT, send the opposite message.
Why are managers the best prompters, and what does that mean for the CEO?
Managers are often the best prompters because good prompting is a management skill: stating intent, giving context and breaking a task into clear steps. Berg describes how Microsoft’s early adopter programme assumed that younger, digitally native staff would write the best prompts. In his words, “It turns out we were dead wrong. Our best prompters are the managers, front-line managers.”
For the CEO, this reframes who should lead adoption. It belongs to the people who already direct work, and the executive team is the most senior group of them. An executive team that does not use AI in its own work withholds the organisation’s most credible demonstration.
The AIPEX view: the executive team’s own AI use is the real AI policy
At AIPEX we see a consistent gap between what organisations write about AI and what their leaders do with it. Staff read behaviour, not policy. A policy that encourages experimentation, issued by executives who never mention their own use, is read as caution.
That makes role-modelling a governance instrument, and it should be managed like one. We recommend that the executive team agree a short, written leadership AI practice, covering three things:
- Disclosure: leaders say when AI contributed to a paper, analysis or decision, and what they checked themselves. This normalises use and models accountability at the same time.
- Boundaries: leaders state, in plain language, which data and decisions are off limits for AI, and hold themselves to it first.
- Learning in public: each executive shares one useful result and one failure each quarter, so that small wins build momentum and mistakes are treated as information.
| What leaders do | What staff conclude |
|---|---|
| Announce a rollout, never mention their own use | AI is for other people; using it is risky |
| Use AI privately and present the output as their own | Use it, but hide it |
| Use AI openly and explain what they checked | Use is expected, and judgement still matters |
| Share a failure and what they learned | It is safe to experiment here |
How does role-modelling connect to return on AI investment?
Role-modelling connects to return because unused licences produce no value, and use follows leadership signals. We examine the gap between buying access and getting value in You Bought the AI Licences. Where Is the Return? Visible executive use is one of the cheapest interventions available: it costs candour, not budget.
It is not sufficient on its own. Role-modelling creates permission; sustained adoption also needs training and redesigned workflows, the subject of From Hype to Habit. Where leaders model use but the organisation cannot move successful experiments into production, the result is enthusiasm without value, the pattern described in AI Pilot Purgatory.
What should the board ask the executive team about leadership AI use?
The board should ask whether the executive team is demonstrating the behaviour it expects of everyone else. Five questions test this:
- How does each member of the executive team use AI in their own work today, and would their teams know?
- Do our board and executive papers disclose where AI contributed, and what was checked by a person?
- What have our leaders said publicly about which data and decisions are off limits for AI?
- What share of our managers have received structured AI training, and who is accountable for closing the gap?
- When did a senior leader last share an AI failure, and how was it received?
If the honest answer is that the leadership team is not yet using AI visibly, that is the place to start. A Fractional Chief AI Officer can coach the executive team in its own use and turn it into a leadership practice the organisation can follow. The Executive Strategy Day gives the leadership team hands-on time with AI against its own decisions, and agrees the disclosure, boundaries and learning norms it will model from the top.
Frequently asked questions
What does it mean for a leader to use AI out loud?
It means saying openly when AI contributed to a paper, meeting or decision, and what the leader checked personally. This signals that AI use is expected at the top while making clear that accountability for the judgement stays with the person. For a full explanation, see Leaders Prompt First: Why AI Adoption Starts in the Boardroom →
Is a written AI acceptable use policy enough to drive adoption?
No. Staff read behaviour, not policy, and a policy that encourages experimentation from executives who never mention their own use is read as caution. BCG found that the share of employees who feel positive about generative AI rises from 15% to 55% with strong leadership support. For a full explanation, see Leaders Prompt First: Why AI Adoption Starts in the Boardroom →
Why do employees stop using AI after an enthusiastic launch?
Visible leadership use creates permission, but it does not sustain habits on its own. Lasting adoption also needs training, redesigned workflows and continued reinforcement once the launch energy fades. For a full explanation, see From Hype to Habit: Making AI Adoption Stick After the Launch →