In short: Leaders don’t need to be able to code AI. They need a reliable sense of what a system can do and where its limits are. This kind of AI literacy for leaders comes mainly from working with the systems yourself, and it decides whether someone actually judges a result or just nods it through.
How deep does a leader need to go into AI on the technical side today? My answer often surprises people: a solid sense of what a system can do right now and where its limits are is enough. That skill determines whether someone actually evaluates a result in the end or just rubber-stamps it.
I see this a lot in workshops. A system delivers a result that sounds convincing, with a clear structure, a confident tone and all the right jargon. That very persuasiveness tempts people to accept it without checking. I once reviewed a market analysis a system had produced and only noticed on the second read that two key figures sounded plausible but were simply made up. Leadership here means reliably telling the difference between a convincing phrase and a correct statement.
An everyday example
In practice
In many of the teams I work with, a human reviews the campaign copy a system suggests before it goes out. The quality of that review depends directly on how well the leader understands how the system arrived at its result, which data it drew on and where it tends to get things wrong. Anyone who can trace that path judges content and tone alike, and that is where the real value of the review lies.
Leading by example means learning yourself
That’s why I’m a big believer in leaders working with the systems themselves, asking their own questions and making their own mistakes. It gives teams permission to do the same.
Setting an example comes from doing the work yourself.
These days I regularly use the same systems as my team so I understand exactly where a result will need rework later.
What this means day to day
In practical terms, that means setting aside one fixed hour every week to work with the systems my team uses daily. No training, no project, just trying things out myself. That hour has taught me more about the limits of these systems than any vendor workshop, and it gives me the footing to stand behind my team’s decisions on the substance.
My answer to the original question: sound judgment is within reach for any leader willing to work with the systems themselves. It’s a bar you can clear within a few months once you really get started, and it shows in every meeting where people suddenly start asking the right questions.
Why is leadership so often the bottleneck?
McKinsey’s Superagency report reaches an uncomfortable conclusion. Three times as many employees use generative AI for at least a third of their work as their leaders assume. At the same time, executives cite employee readiness as a barrier 2.4 times as often as alignment problems within their own leadership. The report concludes that employees are ready and the biggest barrier lies with leadership.
I find this less discouraging than it sounds. If the bottleneck sits with leadership, it can be fixed there too. A leadership team that calls AI important and never uses a system itself sends a message louder than any policy.
What helps teams more than any training?
In the same report, 48 percent of employees name training as the most important factor for adopting generative AI. Almost half of them feel they get only moderate or little support. If you complain that your workforce isn’t on board, first check what you’ve offered them.
Your own example matters at least as much. My book describes a fictional managing director who uses his company’s AI assistant seriously for the first time and asks it to summarize the minutes of the last advisory board meeting. The summary is well written and covers the previous year’s meeting, because he uploaded the wrong file. The next day he tells the story to his leadership team on purpose. A leader who admits to still being clumsy with the new tool makes it easier for everyone else to admit the same.
What should leadership teams review now?
- Clarify who on your leadership team uses AI systems regularly and how the team can tell.
- Measure the share of leaders who work at least once a week with the systems their teams use.
- Talk with your team about an AI output that sounded convincing and was wrong, and about how you could have spotted it.
Frequently asked questions
Do leaders need to be able to code AI?
No. AI literacy for leaders mainly means being able to judge when a result holds up and where a system tends to get things wrong. The fastest way to learn that is working with the systems yourself.
How much time does it take?
Less than many people think. A fixed hour a week with the tools your own team uses does more than a one-off workshop.
Further reading: AI Literacy in the Boardroom: How Advisory and Supervisory Boards Should Build It · How Do Junior Employees Learn When Agentic AI Takes Over Entry-Level Work?
