Please enter a keyword and click the arrow to search the site
Or explore one of the areas below
Turns out the essential human skills are what business schools have been teaching for decades
Critical thinking is more important than ever. AI may seem more objective than we are but it has its own flaws and weaknesses.
Learning the art of questioning will help you enter better prompts. It will also help you interrogate the outputs it gives you.
To really benefit from AI you need to hone your judgement skills around when and how to use it. Make sure you are using it wisely.
In the big flurry of excitement/panic over AI, it’s often stated that organisations need to beef up their human capabilities in order to get the most out of working with the non-humans. So which capabilities are these, exactly, and why do they matter so much?
Empathy is an obvious one, the need to complement artificial intelligence with the EQ that only sentient creatures with a range of lived experiences can bring to the table. But there’s a different skillset that’s just as important, and at the time of writing it’s beyond the grasp of even the brainiest AI. As it happens, it’s exactly what people learn at business school.
“The classic thing in business education is to be a critical thinker, and you need that more than ever with AI,” says Randall S Peterson, Professor of Organisational Behaviour at LBS. “AI is great at gathering information and summarising, but – as the saying goes – garbage in, garbage out. If the information it’s relying on is inaccurate then its output will be too. It’s hopelessly confused out there.”
There’s no AI equivalent of raising an eyebrow at something that just seems “off”. “A human being will just look at it and tell you ‘No, this is wrong’, whereas an AI will tell you that’s how it is, if it’s referenced the wrong thing in some way,” says Peterson. “So we need critical evaluators who can say: ‘Does this make sense?’ Maybe AI will become more reflective, but it isn’t yet.”
Peterson is the author of the book Disaster in the Boardroom: Six Dysfunctions Everyone Should Understand, which outlines the various ways that people can mess thing up when they get together to make decisions as a board. So he’s the first to concede that humans have flaws. Still, he’s horrified to see blue-chip companies incorporating AI board advice.
“AI hallucinations have fabricated facts, misrepresented data and offered incorrect logic,” he pointed out in a recent letter to the FT. “So, while AI can definitely be useful, are we ready to give them a board seat? Definitely not.” He’s adamant that we are not yet at the point where the rewards of using AI outweight the risks. “Is it helpful? Yes. Is it time-saving? Yes. Is it ready to replace humans? No.”
What should organisations be zooming in on, then? “Rather ironically, a lot of what we need is old-school,” says Peterson. Coaching skills such as asking open-ended questions translate into smart prompts for AI, “We need people who can ask better questions, that allow AI to find something you didn’t expect. Are we asking questions that will reveal the things we don’t know, in addition to the things we should?”
He also flags that how an AI is set up and which data it’s trained on makes a difference: who’s programming it, what’s included and not included. “I have a chatbot Randall on my website and it’s trained on everything that I’ve ever written,” he says. “If you ask it something outside of that it will say, “Sorry, I have no information on that.” Some AIs are less modest (as it were). He suggests putting the same prompt into several different AIs to see just how different the outcomes can be: “It’s not as cleanly objective as maybe you’re thinking it is.”
Somewhat alarmingly, it seems we’re overinclined to trust it. “People think that AI is objective, and the thing that drives them nuts about their boss is that they’re not. So somehow it feels better, more appropriate and fair to rely on the AI. I find it scary that people believe that, but it seems to be what’s coming through in the studies.”
Andrew Likierman, Professor of Management Practice and former Dean at LBS has similar concerns about our being overinclined to trust AI. “What comes out of the machine are words on the screen, or rather dots on the screen formed into words. They are not answers. We humans have to understand that impersonal does not mean impartial – those dots are the results of other humans programming, training the data and exercising quality control.”
Likierman emphasises the role of human judgement in our interactions with machines. “In the age of AI, human judgement becomes more important,” he says. “It’s one of the things you’ve got that the machine hasn’t.” Likierman’s book Judgement at Work: Making Better Choices offers suggestions for how to make the right calls more of the time.
When it comes to interacting with AI, he suggests that there are important choices for humans to make, starting with whether to even use AI. Others include: which application of AI? “For example, if it’s a Large Language Model (LLM) should it be CoPilot, Claude ChatGPT or one of the others? Then, what should I ask it? And then, how should I ask it, knowing that the quality of my prompt is crucial to the quality of what I get out as an answer?”
After those preliminary choices comes whether to use AI to check a human judgement or whether to ask the machine first and then assess what it says. Likierman suggests that the order should follow which of the human and machine is likely to have the best answer. “If the machine is going to be best – say on facts which you don’t know, get the machine input first and then assess it. If it’s a complex issue on which you want a second opinion, feed the opinion into the machine and see what comes up.”
He agrees on the need to be mindful that there’s not one thing called AI. “CoPilot is less chatty than ChatGPT, for example. You need to know about the possibilities and limitations of each. Having got past the initial wonderment that this thing produces a magic answer in two seconds flat, it’s about getting used to asking the right questions about what we’re being offered.”
To really benefit from AI, you need to train your judgement as well as your technical skills. This means a focus on the different elements of judgement – knowledge and experience; awareness; trust; feelings and beliefs; choice; and delivery. These are covered in detail in his book.
Knowledge and experience
Understand the relative strengths and weaknesses of the different AI opportunities (LLMs) available.
Recognise that risks of accepting AI advice increase, the less your relevant experience.
Awareness
Be clear of what the machine won’t be aware of – ie the individuals involved and the circumstances right now.
Trust
Is the AI-generated answer consistent with what you know? If not, how good is the reasoning given in explaining the difference?
Test the advice, for example on forecasts, including by cross-checking with independent sources. If necessary, seek corroboration from other trusted sources.
Feelings and beliefs
Are you aware of the biases of the machine, such its tendency to be more positive, and your own biases in interpreting what it suggests?
Is AI’s answer consistent with your values?
Choice
Be clear about why you are choosing one program over another. Are you satisfied with the quality of your prompt? Are additional questions or details necessary?
Have the right number of options been presented, particularly if there are new, highly uncertain or risky choices?
Have you understood the evidence and implications, including risk, of the proposed speed of action, whether speed is of the essence or whether a pause for more facts or reflection would be wise?
Be aware of probabilities, confidence levels and ambiguous words such as ‘possible’ and ‘likely’.
Delivery
What is the track record of you accepting previous AI-generated answers?
How realistic are the delivery assumptions, including the availability of key people and finance, the track record and experience of those involved and the understanding of feasibility and risk?
Have you asked yourself the question: ‘Would I be prepared to defend the choice to colleagues/senior management or in public?’
Is the level of evidence sufficient if a record is needed (e.g. for regulators or to guide colleagues in future), especially if the choice is controversial?
Discover fresh perspectives and research insights from LBS
Get fresh ideas from business experts
LBS in social media
© London Business School 2026
Leave a Reply