AI makes speaking up harder, not easier

A leadership team announces a new set of AI tools on a Monday morning. There is a slide on productivity, a slide on the vendor and a slide on training. The room nods. Nobody asks whether their own role will look the same in two years, whether the tasks they are good at will still be theirs, or whether the rollout is quietly being measured against headcount. Those questions get asked later, in direct messages and over coffee, which is exactly where they do the most damage.

Anxiety that has no obvious owner

Amy Edmondson, Harvard Business School professor and the researcher who has done more than anyone to put psychological safety on the management map, makes a pointed observation in her interview with UNSW Business School's Karin Sanders. The uncertainty around AI (what happens to jobs, which tasks change, how fast) creates a generalised anxiety. It attaches to no single manager or decision, so it hangs over everything. Left unaddressed, it can undermine psychological safety in teams that were previously doing fine.

That matters because psychological safety is team-based. It is the shared belief that people can raise a concern, ask a naive question or admit a mistake without being punished for it. When the background mood is fear about one's future, the calculation each person makes before speaking up shifts. Asking how the new tool works feels like admitting you are behind. Questioning the rollout feels like volunteering to be first on the list. So people stay quiet, and quiet teams are the ones that miss the faulty output, the bad assumption and the better idea.

Make the uncertainty discussable

Edmondson's answer is plain. The only way to strengthen psychological safety is to make it discussable. She advises leaders to stay away from overly negative views of AI and overly positive ones, and to talk honestly about uncertainty. The honest position is that organisations really do not know where this is heading, and that the technology is changing faster than any before it. A leader who says so describes the situation accurately and gives everyone else permission to do the same.

That takes humility, which is in short supply at the top of most organisations. It means encouraging people to discuss the implications openly, express concerns and raise questions, then working through those implications together rather than leaving each person to worry in isolation. The cheerleading deck and the doom memo fail for the same reason. Both close the conversation by pretending to know the answer.

What this means for L&D

Many AI capability programmes are built around tools and prompts. That is useful, but it assumes people will tell you when the tools are not working for them, and that assumption depends on the climate inside each team. If you run learning, innovation or HR, I would add a layer before the next rollout. Give every team a standing slot to name what they do not know about AI and what worries them about it. Ask leaders to go first and to say plainly where their own understanding runs out. Then treat what comes up as input to the plan.

If you want a structured way to build that habit, our psychological safety workshop is designed for team leaders who need to start these conversations. The question worth asking this week: the last time your team heard about AI, what did they want to say that they did not?

Sources: UNSW BusinessThink, interview with Amy Edmondson, https://businessthink.unsw.edu.au/articles/psychological-safety-amy-edmondson-remote-work-artificial-intelligence