Leading AI teams starts with asking better questions
AI leadership is less about having all the answers than creating conditions for better ones. Recent research suggests stronger leaders question more, listen longer and shape clearer turn-taking. For managers, the practical task is to govern AI use, probe assumptions and build disciplined team dialogue.
As AI becomes embedded in everyday work, many leaders are tempted to respond with certainty: clearer policies, firmer decisions and sharper claims about what the technology will do. Yet the better leadership move may be to ask more precise questions. Research from Harvard Business School suggests that when people led AI teams, stronger performance was associated with more questioning, more conversational turn-taking and a more inclusive ‘we’ orientation 23. Stanford’s AI Index also points to a broader context of rapid capability growth, uneven responsible AI practice and widening gaps between expert and public expectations 1. In that setting, leaders need less performative expertise and more disciplined inquiry.
Why questions matter more in AI settings
The Harvard study is useful because it does not treat leadership as charisma or volume. In a controlled laboratory task, leaders worked with both human followers and AI agents, and those who performed well in one setting tended to perform well in the other 2. The practical lesson is narrow but important: leadership skill transferred when people could gather information, test assumptions and integrate fragmented clues. That does not mean AI teams are the same as human teams. It does suggest that the habit of asking well-formed questions is a stable leadership behaviour across different kinds of collaboration.
Leaders often assume they must sound decisive to be effective. In AI work, that instinct can be counterproductive. Models can produce fluent answers quickly, but fluency is not the same as accuracy, context or fit for purpose. Better questions help surface what the system is missing, what the team has not defined and where human judgement still matters. The research does not prove that questions alone cause success, but it does show that stronger leaders created more back-and-forth and learned more from the exchange 3.
Use questioning as a management discipline
A useful practice is to make questioning a formal part of AI decision-making. Before a team adopts a tool or approves a use case, ask: What problem are we solving, what would better look like, what could go wrong and who needs to be involved early? That approach fits MIT Sloan’s warning that broadly applicable AI tools can create risks around data, bias, misleading outputs and hidden costs, while tailored solutions bring their own governance and vendor dependency issues 3. Questions are not a substitute for controls, but they help leaders choose the right controls.
Another practical step is to separate exploratory questions from approval questions. Early-stage conversations should invite curiosity: what do we notice, what patterns are emerging, what assumptions are we making? Later-stage conversations should become more exacting: what evidence supports this, how will we monitor drift, what happens if the model is wrong? Stanford’s AI Index notes that capability progress is outpacing responsible AI reporting in many settings, so leaders should expect uncertainty rather than assume it away 1. A questioning culture makes that uncertainty discussable instead of hidden.
What strong AI leaders ask their teams
The Harvard findings imply that good AI leadership is interactive, not broadcast. Leaders asked more questions and teams showed more conversational turn-taking 3. In practice, that means avoiding one-way briefings and replacing them with structured dialogue. Useful prompts include: What information do we trust most? What is the model optimising for? Where might it be overconfident? What decision remains ours, even if the tool is useful? These questions slow the conversation just enough to improve it.
It also helps to ask who is missing from the room. AI work often sits across operations, technology, legal, people and risk functions. MIT Sloan’s management guidance emphasises cross-functional governance, clear usage rules and training that teaches employees how to instruct and interrogate AI tools 3. That is not bureaucratic overhead; it is how organisations reduce blind spots. When teams are encouraged to ask questions across boundaries, they are more likely to spot data issues, policy gaps and real-world friction before those problems spread.
Build a team environment where questions are welcome
Leaders shape whether people feel safe to question an AI recommendation. If the tone is defensive or overly technical, team members may stay quiet even when they see problems. The Harvard article notes that warm, positive language mattered more with human followers than AI ones, which is a reminder that people still need respect, clarity and room to think 3. In AI projects, this means rewarding the person who asks the awkward but necessary question, not only the person who produces the neatest slide.
A practical routine is to end AI meetings with three checks: what is still uncertain, what would change our mind and what do we need to test next? Those questions keep the work moving without pretending it is settled. They also reduce the risk of premature confidence, which can be costly when a tool is adopted too broadly or too quickly. Our advisory view is that leaders who normalise well-timed challenge usually make better use of AI because they are less likely to confuse speed with sound judgement.
From better questions to better governance
As AI adoption becomes routine, the leader’s role shifts from explaining the technology in general terms to governing its use in specific contexts. Stanford’s report shows how quickly AI is spreading, but it also highlights uneven safety practice and a growing gap between technical progress and responsible oversight 1. That gap cannot be closed by a policy document alone. It closes when leaders keep asking what the tool is doing, what it is not doing and how people will know the difference.
For executives, the challenge is not to become the person with every answer. It is to create an organisation that asks disciplined questions early, listens carefully and acts on what it learns. In AI teams, that habit is not a soft skill. It is a core control.
Sources
- The 2026 AI Index ReportStanford University — Human-Centered AI · University research · link checked 15 October 2026
- Think You’re a Good Leader? Try Leading an AI TeamHarvard Business School AI Institute · University research · link checked 15 October 2026
- How to manage two types of generative artificial intelligenceMIT Sloan School of Management · University research · link checked 15 October 2026
- Who should be the head of generative AI — and what they should doMIT Sloan School of Management · University research · link checked 15 October 2026
Further professional reading
External professional resources, separate from the numbered sources above. These are not university studies, and listing them does not indicate any collaboration, certification, affiliation or endorsement.
- Global Institute of Directors (opens in a new window)External professional resource · governance and directorship
- World Coaching Institute (opens in a new window)External professional resource · coaching practice