Leadership in the age of AI: Why human discernment is essential

Almost every leader we speak to has multiple AI questions on their desk, and few of them are really about the technology. They are about people, trust and working well with the colleagues who hold the technical keys. We are reminded by the leaders we speak to how much leadership has changed: Not long ago, AI was a topic for the technology function, and today it sits cross-functionally in every boardroom conversation.

The leadership press reflects it. Korn Ferry's latest This Week in Leadership puts AI alongside the cost of returning to the office and the need to "unlearn" old ways of working, while IMD is publishing on HR leading the "hybrid human-machine workforce" and on the Chief Legal Officer as architect of enterprise AI governance. The common thread is that AI is no longer one function's project, and the tools are only as good as the discernment of the people using them. Here are the four questions we hear most, particularly from leaders outside technology.

Four key Questions

How should leaders use AI themselves?

Start by using it, in a small and honest way. It is hard to lead a conversation about something you have only read about. A few habits seem to separate leaders who get real value from those who get a lot of noise:

1. Use it as a sparring partner, not an oracle. Research from Monideepa Tarafdar, published by Berkeley's Greater Good Science Center this month, found that knowledge workers who deliberately used generative AI to challenge their own ideas improved their performance. Those who simply accepted fast, confident answers tended to get speedy but lower-quality solutions. Asking "what is the strongest argument against this?" is a good place to begin.
2. Be open about how you use it. If you draft a first pass of a board paper with AI, say so, and say what you checked. Teams copy what leaders do far more than what they say.
3. Keep the decisions that matter. Use AI to widen the options, then apply your own judgement. IMD's Gerd Gigerenzer makes a related point about decisions in uncertain times: good judgement is often about knowing which tool fits the problem, not reaching for the most complex one.
4. Protect your time for people. The most useful thing AI can give a leader is time for the conversations that only they can have.

How should leaders encourage their teams to use AI?

Most people are neither wildly enthusiastic nor stubbornly resistant. They are curious, a little anxious, and unsure what is allowed. Clearing up ambiguity and removing hesitation is where leaders can help most:

1. Give clear permission and clear boundaries. People will experiment if they know where the lines are. Without that clarity they either avoid AI entirely or use unapproved tools quietly, which creates its own risks.
2. Start with real problems. Pick a handful of everyday frustrations in the team (reporting, research, first drafts, meeting follow-ups) and test AI there. Small wins build confidence better than big announcements.
3. Be clear about which kind of AI investment you are making. At the recent Technology Chapter Summit UK at Ashridge House, one useful framing was that AI investment tends to fall into three tiers: generalist tools for everyone, product-level capability, and internal use cases. Knowing which tier a request belongs to helps teams (and leaders) set realistic expectations on cost, risk and return.
4. Ask for benefits, not just activity. The same discussions kept returning to benefits realisation. Usage numbers are not outcomes. Ask what has actually changed: time saved, quality improved, decisions made faster.
5. Make room to "unlearn". Korn Ferry notes that leaders increasingly look for people who can let go of old ways of working and move quickly to what comes next. That is easier when leaders say openly that some of what made people successful yesterday may need to change.
6. Talk about the job question honestly. People are wondering what AI means for their role. Silence tends to be filled with worst-case assumptions. Even "we don't know yet, and here is how we'll decide" is better than saying nothing.

How should non-technical leaders partner with CIOs, CISOs and CTOs?

This is the question we hear most from peers outside technology, and it has a reassuring answer: you do not need to become technical. You need to become a good partner.

Deloitte's 2026 Global Technology Leadership Study, which draws on more than 660 senior technology leaders, suggests that no single function can manage AI risk end to end. It cuts across cyber, data, operations, compliance, suppliers, finance and the business itself. Deloitte also makes the point that shared responsibility should not become ambiguous accountability. Somebody still needs to own each decision.

In practice, this might look like:

1. Bring the business problem, not the tool request. "We lose three days a month reconciling these reports" gives your CIO and CTO far more to work with than "we'd like to buy this product".
2. Ask a few simple questions, and expect straight answers.
   What data does this use, and where does it go?
   What happens if it gets something wrong, and who finds out?
   Who owns this once it is live?
   How will we know it is delivering the benefit we expected?
3. Don't let technology leaders be cast as the blockers. Technology teams often tell us they are seen as the department of "no". Early involvement turns that around. Bring your CISO into the conversation before a pilot launches, not after, and they can usually help you find a safe route to yes.
4. Recognise that old assumptions have gone. Shadow AI is the new version of the BYOD (bring your own device): People will use what helps them. A secure, approved option that is genuinely good is more effective than a ban.
5. Widen the circle. IMD's piece on the Chief Legal Officer describes the CLO moving into the role of architect of enterprise AI governance. IBM's recent study found that nearly half of organisations do not involve the CHRO when AI strategy is being defined. Legal, HR, finance and the business all have a stake, and regulation is moving too, including the EU AI Act, whose main obligations are reaching enforcement this year.

How do we make sure critical thinking isn't degraded over time?

This may be the quietest risk, and probably the most important one.

The IBM Institute for Business Value surveyed 1,500 CHROs and 8,800 employees. It found that 60% of employees worry AI is eroding their skills, with critical thinking cited most often as the one declining. There is also a gap in perception: 71% of CHROs see the ability to supervise, validate and override AI outputs as the workforce's most essential skill, while only 29% of employees rank judgement as important.

BCG describes a related risk as "cognitive debt", the gradual erosion of critical thinking, judgement, curiosity and originality when AI substitutes for thinking rather than supporting it. Its research points to a particular symptom: reduced ownership. More than half of the leaders it spoke to cited it. One put it plainly: "I can see employees' sense of ownership going down. If the outcome is not positive, the responsibility is laid on AI."

None of this is a case for stepping back from AI. It is a case for designing how we use it. Some practical ideas:

1. Make the human accountable, always. "The AI suggested it" should never be an acceptable explanation. The person who uses an output owns it.
2. Ask people to show their thinking. When reviewing work, ask what they checked, what they changed, and what they disagreed with. It signals that judgement is valued.
3. Use AI to challenge, not just to answer. Build the Tarafdar approach into team routines: ask the tool for counter-arguments, risks and blind spots before a decision, not a confirmation afterwards.
4. Keep some work human on purpose. Just as pilots still practise manual flying, teams may want to keep certain skills (analysis, writing, negotiating) regularly exercised without assistance.
5. Train the supervisory skill. Reviewing AI output, spotting errors and knowing when to override are skills in themselves. They need practice and recognition, and they should show up in development plans.
6. Check in on how people feel. If 60% of employees are worried about their own skills, it is better to talk about it than to find out later.

Continuing the conversation

Nothing here is a finished answer. The leaders we speak to tell us they are working it out as they go, comparing notes, and borrowing what works from one another. That is partly why it matters to have the right people to talk to, in the right setting, over time.

If you are a leader grappling with these questions, we would like to hear how you are approaching them. The conversation is much better when it is shared. Please follow us on LinkedIn or enquire about joining The Leadersphere to meet with senior leaders facing similar challenges. 

Sources

1. IBM Institute for Business Value, New IBM CHRO Study: AI Puts Critical Thinking at the Center of Workforce Priorities (21 Sept 2026): https://newsroom.ibm.com/2026-09-21-new-ibm-chro-study-ai-puts-critical-thinking-at-the-center-of-workforce-priorities
2. BCG, When Everyone Uses AI, Companies Risk Critical Skills (2026): https://www.bcg.com/publications/2026/when-everyone-uses-ai-companies-risk-critical-skills
3. Monideepa Tarafdar, How Artificial Intelligence Can Help With Critical Thinking, Greater Good Science Center (12 Oct 2026): https://greatergood.berkeley.edu/article/item/want_to_use_ai_to_improve_your_work_have_it_disagree_withyou
4. Deloitte Insights, How AI is changing the CISO role: https://www.deloitte.com/us/en/insights/topics/technology-management/ai-changing-ciso-role.html
5. Korn Ferry, This Week in Leadership (Oct 12 to Oct 18): https://www.kornferry.com/insights/this-week-in-leadership/twil-oct16-2026
6. IMD (I by IMD), Leadership: https://www.imd.org/ibyimd/tag/leadership/ (including HR needs to start owning how work gets done, AI and the Chief Legal Officer, and Decision making in uncertain times)
7. CISO Forum, session on AI risk governance (NIST AI RMF, EU AI Act, ISO/IEC 42001): https://www.cisoforum.com/sessions

Photo by Omar:. Lopez-Rincon on Unsplash