As AI agents take over more of the routine decisions inside companies, it's tempting to conclude leadership will become easier. I believe the opposite is true.
Delegate the easy decisions and what's left on the leader's desk are, by definition, the difficult ones: ambiguous, unquantifiable, with no clean precedent.
These are decisions where even excellent judgment will often turn out to be wrong. Machines will increasingly take the decisions that are easy to get right. Yet for every important decision, we still want a human who is accountable for it. The job of a leader will become more challenging and more important at the same time.
Why more mistakes will be visible
I expect leaders’ errors to become more visible, for three reasons.
The first is the change in decision portfolio. As easy decisions leave the human portfolio, the share of human decisions that go wrong rises, even if leaders are no less capable than before.
The second is salience of human mistakes. When automated systems handle routine work consistently and at scale, a human misjudgement stands out against a backdrop of near-flawless execution.
The third is traceability. In the age of AI, there will be an increased demand for human accountability. The question 'who decided this?' will have a documented answer, and for hard decisions, it will not be the algorithm.
What accountability looks like in practice
To illustrate this point, take a recent example from my own experience. I commended a colleague for a rigorous market expansion report, which they acknowledged was drafted using AI. I told the colleague that as long as they stand by every word in the report, it's not important whether the report is written by themselves, human collaborators or AI agents.
In this sense, it's not very different from a CEO signing off on financial statements and being accountable for every number —even if these numbers were prepared and checked by others.
Unfortunately, in today’s environment, some business leaders lose sight of the vital role of accountability. Not checking what AI says in your name may undermine trust in you and your standing in the organisation.
What founders and investors should do differently
First, decide what you are optimising for before you automate, and be explicit about it. A stated objective helps turning a later change into learning rather than face-saving.
Second, name the accountable human for every consequential AI-assisted decision. 'AI recommended it' is not an answer, and your customers, employees and regulators won't accept it.
Third, tell your teams in advance that bolder decisions will sometimes fail. Leaders must respond to transparency with transparency, and that means honestly communicating the risks and the costs of potential mistakes before they happen.
Fourth, and this is for investors in particular: reward the quick, honest reversal. In the current complex, uncertain and evolving environment, learning and adapting is inevitable and helpful. A board that treats a public U-turn as weakness will get leaders who conceal their errors until they are too large to hide.
The machines will take the easy calls. What remains will be harder, and we will get more of it wrong. The leaders who thrive will be those who know and own these rules of the game.
