McKinsey’s Organization practice published a blog post this month with a title that should sting a little: you cannot lead AI from the sidelines. Their argument, drawn from more than two dozen AI transformations, is that AI adoption is now outpacing leadership readiness, and that the leaders creating the most value are the ones using the technology themselves, not the ones being briefed on it. I think this is the most important leadership point of the year, and I want to add what it looks like from the inside.
The pattern, and why it sounds familiar
The McKinsey team describes a bottleneck that has quietly moved. It is no longer technical feasibility. It is whether leadership understands AI well enough to redesign work, build trust and capture value. One quote from their research deserves to be printed out: the worst thing you can do as a role model is to explain AI without having used it yourself[1]. For years, executives could lead technology programmes by asking sharp questions and sponsoring budgets. Agentic AI ends that model, because the unit of work is no longer a tool you approve. It is an evolving workflow you have to understand.
Fluency comes from doing
The strongest examples in the McKinsey piece are almost embarrassingly simple. Leadership immersion sessions where executives pick a pain point from their own lives and build an agent to solve it, not for the use case, but for the moment when AI stops being a concept and becomes a collaborator. A global bank putting senior leaders through hands-on training and watching them leave energised. Reverse mentorship, where digitally fluent colleagues who are junior in hierarchy but senior in practice teach upward[1].
The workforce data says this is exactly where the leverage sits. Gallup’s 2026 State of the Global Workplace research finds employees whose manager actively supports AI use are more than twice as likely to use it frequently, and 8.7 times more likely to say AI has transformed their work. Yet only about three in ten employees say their manager provides that support[2]. Leadership behaviour is the single strongest activation lever we know of, and it is mostly unused.
Managing human-agent teams is a new job
The second half of the McKinsey argument matters just as much. As agents take over parts of workflows, managers stop supervising people who produce work and start orchestrating output from humans, agents and systems, at higher speed and more variable quality. Their researchers call the new burden discernment: is this answer right, can the team defend the logic, where does the risk sit. One practitioner puts it bluntly: teams will bring leaders AI slop, and managers must know how to spot it and how to coach the person who produced it[1]. I would add: this discernment cannot be delegated to a governance framework. It has to live in the operating rhythm. One example from their work is a weekly review with a rotating challenger role whose job is to find evidence against the leading conclusion before the decision is made[1]. That is decision hygiene, and it is exactly the discipline most organisations skipped even before AI.
The rest of the research agrees
This is not one firm’s opinion. Deloitte’s survey of 3,235 enterprise leaders finds close to three quarters of companies planning to deploy agentic AI within two years, while only 21 percent have a mature model for governing what those agents do[3]. You cannot govern what you have never used. And Stanford’s 2026 AI Index shows organisational adoption at 88 percent while agent deployment stays in the single digits across most functions[4], a gap that sits exactly where leadership fluency should be. Different methods, same conclusion: the constraint has moved from the technology to the people directing it.
What I see in the field
In a DAX headquarters environment and in conversations with mid-sized owners, I see the same tell McKinsey describes: leadership teams that have been briefed on AI a dozen times and have used it roughly never. You can hear the difference in one meeting. Leaders who have built something, even one small agent for a private pain point: ask about workflows, failure modes and ownership. Leaders who have only been briefed ask for another briefing. The credibility gap is real, and teams read it instantly.
What this means for you
Three moves, none of them expensive. First, build one agent yourself on a problem you personally own, the point is the aha, not the output. Second, make AI fluency a stated management expectation, not an optional module; the Gallup numbers say your managers are the multiplier. Third, put discernment into your rhythm: one weekly review, one rotating challenger, one simple log of where AI was right and where human judgment saved you.
Building exactly this hands-on fluency, for leadership teams and for the people who do the work: is what our AI Foundations Workshop exists for, and keeping your judgment sharp as the technology moves is what the AI Decision Partner is built around. If your organisation is being briefed instead of building, that is fixable in weeks.
Personalised answer within 48 hours.
Sources
- McKinsey & Company, People & Organization Blog: You Can’t Lead AI From the Sidelines (Metakis, Srinivasan, Catalino, James), 7 July 2026. mckinsey.com
- Gallup, State of the Global Workplace 2026: The Human Side of the AI Revolution. gallup.com
- Deloitte, The State of AI in the Enterprise 2026 (survey of 3,235 leaders). deloitte.com
- Stanford HAI, The 2026 AI Index Report. hai.stanford.edu
The AI Act Just Moved. Your Deadline Did Not.