Do you need to be technical to lead AI?

The most common reason senior people give for handing AI to the technologists is that they do not understand the technology. It is an honest reason and the wrong one, because the decisions that belong to a leader are not technical decisions. Whether a credit decision may be executed by a model or only recommended by one is a question about accountability and appetite. Whether the organisation keeps some underwriting unaided so that its underwriters can still check the model is a question about capability. Who has the authority to stop an agent, and how often a human reviewing machine output actually reaches a different answer, are questions about governance. A leader who could not build any of these systems can decide all of these things. A leader who could build them has no particular advantage in deciding them well, and some disadvantage, because the builder's instinct is to trust the build.

What technical fluency is for#

Fluency helps and should be worked at, and the reason is precise. A leader who has used the tools knows three things a briefing cannot teach: how confident the output sounds when it is wrong, how much of the work it did without being asked, and how quickly one stops checking. Those three are the failure modes of oversight, and knowing them from the inside is what makes a leader's questions sharp. That is a different thing from being able to build the system, and the two are regularly confused, usually by people who would prefer the leader to stop asking.

The regulation makes the same distinction. Article 4 of the EU AI Act, in force since February 2025, requires providers and deployers to ensure 'a sufficient level of AI literacy' among the staff dealing with their systems. Literacy there means enough understanding of what a system does and where it fails to use it responsibly. It does not mean engineering, and a leadership team that reads it as engineering has found a reason to delegate a duty that was written for them. The fuller account is at what AI literacy means for leaders.

What cannot be delegated is the judging#

Banking's model risk regime, the most mature oversight practice any industry has for machine-produced numbers, rests on effective challenge by people with three things: independence, standing and expertise. Independence is structural and can be arranged. Standing is political and can be granted. Expertise is built by doing the work. An organisation that automates the work its challengers learned on removes the third leg while keeping the other two, and its governance chart looks the same afterwards. That is Lisanne Bainbridge's 1983 argument about process control, applied to knowledge work, and it means the leader's job includes keeping people who can still challenge, as well as holding the authority to.

Vaccaro, Almaatouq and Malone's meta-analysis of 106 experiments found that human and AI combinations performed worse on average than the better of the two alone, with the losses concentrated in decision-making. The benchmark is an oracle nobody has in advance, so the result does not say that oversight is useless. It says that putting a human next to a machine is not oversight until the human can tell when to disagree, which is a capability question and therefore a leadership one.

What happens to leaders who delegate it#

The record of AI programmes that failed is a record of decisions nobody senior made. RAND's 2024 interviews with 65 practitioners list leadership-driven failure first among the root causes: misunderstanding what problem the project was for, or optimising the wrong metric. Gartner's three reasons for the cancellations it forecasts are escalating costs, unclear business value and inadequate risk controls, each a decision that should have preceded the purchase. Klarna's chief executive said in May 2025 that cost had been too dominant a factor in a decision he had made. None of these was a technical failure and none would have been prevented by a more technical leader. Each would have been prevented by a leader who kept the decision.

McKinsey's 2025 survey found the chief executive's personal oversight of AI governance to be one of the attributes most associated with reported profit from generative AI, and found that 28 per cent of organisations had it. The correlation is weak and the outcome self-reported, but the direction is the one this page argues: the money follows the leader who owns the rules, not the one who understands the model.

What nobody has measured#

No study compares technically trained leaders with the rest on AI outcomes, and this page would be surprised by a large effect either way. The evidence here is on where failures come from, which is consistently the unmade decision, and on what oversight requires, which is the capacity to disagree. Whether a computer science degree helps with either is untested, and the argument from the shape of the decisions is the argument.

What a non-technical leader does on Monday#

Use the tools for a week, on real work, and notice the three failure modes. Write the list of decisions the organisation makes and mark each inform, recommend, execute or never. Name who owns each. Ask the technologists one question they cannot answer with a demonstration: how often does a human reviewing this system's output reach a different answer, and what happens when they do? Then decide what the organisation keeps practising unaided. None of that requires code. All of it requires the leader to stay in the room. The full argument is at AI leadership.

Key sources

• Vaccaro, M., Almaatouq, A. and Malone, T. (2024). When combinations of humans and AI are useful. Nature Human Behaviour. Graded entry.
• Bainbridge, L. (1983). Ironies of automation. Automatica. Graded entry.
• Ryseff, J., De Bruhl, B. F. and Newberry, S. J. (2024). The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed. RAND. Graded entry.
• Gartner (2025). Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027. Graded entry.
• Singla, A. et al. (2025). The state of AI. McKinsey. Graded entry.
• Siemiatkowski, S. (2025), Klarna, reported by CX Dive from a Bloomberg interview. Graded entry.

Related SuperSkills research#

The definition this page rests on is at AI leadership. On the literacy duty, what AI literacy means for leaders. On what a human next to a machine has to be able to do before it counts, meaningful human oversight. On what a chief executive should understand personally, what board oversight of AI looks like.

About this research#

Rahim Hirji is the author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. He has run, grown, bought and advised businesses with AI in them. Findings are attributed to the studies and statements that produced them and kept separate from the interpretation. This is a living reference, reviewed and updated as significant new evidence appears.

How this research works  ·  Reviewed quarterly  ·  Found an error? Tell me and it is corrected on the page.