There is a question I would like more leaders to consider before introducing another AI tool: how easy would it be for someone on your team to admit they do not understand how to use it? I am particularly interested in the experienced colleague whom everyone turns to for advice. If that person needed help with something others seemed to grasp quickly, would they feel comfortable asking in front of the team?
Someone recently shared with me that they felt uncomfortable during an AI demonstration at work. They had lost track of part of the process, but their colleagues appeared to be following along and they hesitated to interrupt. As an experienced member of the team, they were concerned that asking for another explanation might make them look less capable. They left intending to work it out on their own.
What struck me about their account was how easily the presenter could have mistaken their silence for understanding. They wanted to learn, but asking for help felt like risking their professional reputation. It made me think about how much we expect people to adapt without considering whether they feel comfortable letting others see where they are struggling.
The discomfort of becoming a beginner again
Learning something unfamiliar can be particularly awkward when being knowledgeable has become part of your professional identity. Someone may have spent years developing sound judgment and earning colleagues’ confidence, only to find themselves struggling with a tool another person seems to use effortlessly. Their experience still matters, but it may be hard to remember in a meeting where speed with the technology draws most of the attention.
I would be cautious about describing that hesitation as resistance. Before deciding someone lacks an adaptable mindset, a manager should explore what makes participation difficult. The person may need a slower explanation, an opportunity to practise or clarity about when the tool is appropriate. They may also have a legitimate concern that has been overlooked in the enthusiasm to get everyone using it.
Leaders can make this harder without intending to. Remarks about how easy a tool is, or how everyone should already be using it, leave little room for someone having a different experience. A more useful introduction would acknowledge that people have different starting points and explain what support will be available. That gives employees something concrete to work with, rather than leaving them to wonder whether asking for help will affect how they are viewed.
What psychological safety means while learning
Psychological safety concerns whether people believe they can take interpersonal risks within their team, including asking questions, acknowledging uncertainty and raising concerns. In her research on psychological safety and team learning, Amy Edmondson found an association between psychological safety and learning behaviour across 51 work teams in a manufacturing company. The study predates generative AI, but the question it addresses is relevant: what allows people to make the difficulties of learning visible to one another?
Applied to AI, I want a team member to feel able to say they cannot explain an output or don’t know how to check it. That admission gives the team an opportunity to investigate before relying on the result. It also helps a manager distinguish between someone who can produce an answer and someone who understands enough to use it responsibly.
This does not remove the expectation that people develop competence. A leader still needs to define what good work requires and address performance gaps. However, how they respond to a request for help should make those gaps easier to identify early. Humiliating someone for not knowing leaves the underlying difficulty unresolved and gives them a reason to conceal it next time.
Training needs room in the working day
An invitation to learn has limited value when the workload leaves no time to do it. Imagine an employee who is encouraged to experiment with AI but is still expected to complete every existing task at the same pace. Practising, checking mistakes and asking a colleague for guidance all take time. Without some adjustment, the employee may end up trying to learn after work or avoiding the tool until a deadline makes its use unavoidable.
The OECD’s surveys of employers and workers in manufacturing and finance across seven countries found that training and worker consultation were associated with better outcomes for workers. These findings do not establish that any particular training programme will succeed, but they support giving employees preparation and a voice in how AI enters their work.
For a manager, that means being specific about where learning fits. If the team is expected to spend time practising a new process, what will be postponed, reduced or reassigned? It is also worth asking employees which parts of the work would benefit from support. A demonstration can show what a tool can do, while a conversation with the people doing the job can reveal what they actually need to learn.
Leaders can show their own learning process
A leader does not have to perform uncertainty or pretend to know less than they do. There is usually enough genuine learning to discuss, and these conversations can include the whole team. I ask my team members to share examples of AI responses they initially found convincing, explain what prompted them to check the information and describe what they discovered. Discussing these experiences together makes the checking process visible and gives everyone a more realistic understanding of what competent AI use involves.
I find that more useful than a presentation containing only successful examples. If employees see the finished result without the unsuccessful attempts, they have little basis for judging whether their own difficulties are ordinary. Discussing a problem and how it was resolved gives them something they can apply when they encounter a similar situation.
The response to someone else’s difficulty matters just as much. When a colleague admits they are stuck, taking the time to understand where they lost confidence is more helpful than immediately demonstrating how quickly you can do the task. The aim is for them to leave better able to work independently, with enough confidence to ask again when they need to.
Three ways to make learning easier this week
Make one question easier to ask. At your next team meeting, identify a part of the AI workflow that deserves closer examination, such as checking whether a generated summary is accurate. Ask where people are finding that step difficult, and offer a private way to respond as well. This is more specific than asking whether everyone is comfortable with AI, and it gives you a practical starting point for support.
Create a small, protected practice session. Set aside twenty minutes to work through one relevant task using an approved tool and information suitable for that setting. Agree what can wait during that time, and focus on understanding the process rather than comparing who finishes first. End by discussing what still needs explaining, so the session informs the next step in learning.
Acknowledge a useful admission of uncertainty. When someone flags an output they cannot verify or asks for help before relying on it, explain why that was a responsible decision. Follow through by helping them resolve the issue or finding someone who can. The team then has a concrete example of what happens when a person speaks up.
I would judge progress partly by the quality of the questions people become willing to ask. A team that begins discussing its difficulties may appear less confident than one in which everyone nods through a demonstration, but the manager has a much clearer view of where support is needed. If we want people to keep learning throughout their careers, we need to give them a reasonable way to be inexperienced at something without treating their wider competence as if it has disappeared.