All insights
    AI Literacy4 min read

    Keep your judgement when AI has a view

    By Sascha Cutura

    A closed charcoal laptop beside an open cream notebook and a brass pen

    A few years ago, asking a computer for advice meant running a report and interpreting it yourself. Today you can ask an AI assistant whether to enter a market, restructure a team or delay a launch, and within seconds you will get a fluent, well-organised answer with a recommendation at the end.

    That answer will sound more certain than most of your advisers. It will rarely say "I don't know". And because it arrives so quickly and reads so well, it is very easy to treat it as more considered than it is.

    These tools are useful sparring partners, and leaders should get comfortable with them. The question is what happens to judgement when the sparring partner is always available, always confident and often rather agreeable.

    Fluent is easily mistaken for right

    Language models are built to produce plausible text. Plausible and correct overlap a great deal, which is why the tools are useful. They are not the same thing, and the gap tends to appear exactly where leaders most need help: in situations that are new, ambiguous or specific to their own organisation.

    There is a second effect that gets less attention. These tools tend to work with the framing you give them. Ask "why is this acquisition a good idea?" and you will get a persuasive case for the acquisition. Ask the opposite and you will get a persuasive case against it. If you are already leaning one way, the tool will very often help you lean further.

    Used carelessly, it confirms what you already think. Used well, it can show you what you have missed.

    Three habits that keep the judgement with you

    The executives who use AI well have a few habits in common. None of them are technical.

    They ask for the case against. Before acting on a recommendation, they ask the tool to argue the opposite position as strongly as it can, then to list what would have to be true for the original recommendation to fail. This takes two minutes and regularly surfaces a risk nobody in the room had raised.

    They separate facts from framing. An AI answer mixes things it can reasonably know with assumptions it has filled in. Good leaders pull those apart. Which figures came from our own data? Which are general claims? Which are guesses dressed as facts? Anything that matters gets checked against a source a person can stand behind.

    They decide, at least roughly, before they ask. Before consulting the tool on a significant decision, they write down their own instinct and the reasons for it, even in a couple of lines. Then they compare. If the AI changes their mind, they can see exactly why. If it does not, they have at least tested their own reasoning.

    Write down your own view first. Then let the machine argue with it.

    Accountability stays where it was

    Whatever role AI plays in a decision, the accountability stays exactly where it was. A board will not accept "the model recommended it" as an explanation, and neither will a regulator, a customer or a member of staff whose job has changed as a result.

    This is worth saying out loud in leadership teams, because the pressure runs the other way. When a tool produces a clear recommendation, disagreeing with it can feel like stubbornness. It helps to agree, explicitly, that AI input is one voice in the room and that the people around the table are still expected to own the outcome.

    Individual habits are a good start. They hold up better when the whole leadership team shares them. A simple rule for board and executive papers works well: if AI contributed to the analysis, say where, and say what was checked. It costs a paragraph and it changes the conversation from "is this right?" to "how do we know?"

    This is also why AI literacy for leaders is different from tool training for teams. Teams need to know how to use a system safely in their daily work. Leaders need the additional habit of treating a confident answer as the start of a discussion, not the end of one.

    The 90-Day AI Adoption Plan

    A practical guide for leaders, covering people, governance and a first delivery, with a skills plan and checklist.