What it is
Protect psychological safety on purpose while you roll AI out. Psychological safety means people feel able to ask questions, raise problems and admit mistakes without being punished. An AI rollout needs exactly that: staff have to say where the tool is wrong and explain how the work is really done.
What the evidence shows
- AI adoption can lower psychological safety. In a survey that asked the same 381 employees three times, more AI adoption at a company went with lower psychological safety, and then with more symptoms of depression. Where leaders were seen as ethical, the damage was smaller.
- Fear leads to hiding knowledge. In a survey that asked the same 402 employees three times, people who feared losing their job to AI hid more of what they knew from colleagues. Employees who felt confident they could learn AI were less affected.
- The opposite has also been found. In a field experiment at a technology services firm, 42 teams with 316 employees were chosen at random, and some were given an AI assistant built on the firm’s own knowledge. Staff with the tool reported more ties with colleagues for sharing knowledge, not fewer. The answers are self-reported, and the study is a working paper. What the tool is for, and whether people fear for their jobs, may decide which way it goes.
- Consultation and training help. OECD surveys in seven countries found that training workers and consulting them about AI went with better outcomes for workers.
How to do it
- Before the rollout, tell staff plainly what the project is for and what it means for their jobs. If you do not know yet, say so, and say when you will know.
- Consult the people whose work will change before the design is fixed. At Commonwealth Bank of Australia, the union said call volumes were rising when the bank announced it would replace 45 roles with a voice bot. The bank later said its assessment had been an “error”.
- Train people early. Confidence in learning AI softened the fear in the study above.
- Say whether using AI is expected, so people do not hide it. In four experiments with 4,439 people, people who used AI at work were judged less competent and less motivated.
- Make it safe to report AI errors. Thank the person who finds one.
- Have leaders use the tools themselves and talk openly about where the tools fall short.
- Ask about psychological safety in a short survey before the rollout and again some months in.
What to measure
- Psychological safety, using a short published questionnaire, before and during the rollout.
- The number of AI errors and concerns staff report. A rollout with none reported is a warning sign.
- How many staff take the training.
Where it fails
- Reassurance you cannot keep. If jobs will change or go, saying otherwise destroys trust when the truth comes out.
- Surveys with no follow-up. Asking and then doing nothing makes things worse.
By company size
- Medium: the owner or managing director can do most of this in person, in team meetings.
- Mid-market and enterprise: line managers carry the message. Brief them first, and measure by team, not only company-wide.
Assessments for this practice
These are existing questionnaires, tests and guides you can use. Each card gives the question it answers, who answers it and whether you can use it freely. All assessments.
- Where do people feel unsafe to speak up: in the organization, the team or one-to-one? Perception of Psychological Safety scale (PPS) Each employee · 14 questions Free to use and adapt
- How safe does each employee feel at work? Psychological Safety Inventory (PSI) Each employee · 30 questions Ask the authors
- Do people in this team speak up, or stay silent? Composite measure of psychological safety (observation and survey) A trained observer and the team’s members · One observed team meeting, plus a 19-question survey Free to use and adapt
Company cases
Companies that did this
- Duolingo The chief executive wrote to all staff with what would change: contractors, hiring, performance reviews and headcount. He also wrote that it “isn’t about replacing” employees, and promised training.
- Gilberts Chartered Accountants Partners and managers use the tools and share what they find, so that staff “feel safe to try new tools but are clear on what is acceptable”.
- META-Regalbau From the start, logistics managers, the works council and all employees were told what the analysis was for and how the technology worked.
Sources
- Kim, Kim and Lee (2025). The dark side of artificial intelligence adoption: linking artificial intelligence adoption to employee depression via psychological safety and ethical leadership. Humanities and Social Sciences Communications, 12(1). What this study did and found
- Kim and Kim (2024). How artificial intelligence-induced job insecurity shapes knowledge dynamics: the mitigating role of artificial intelligence self-efficacy. Journal of Innovation & Knowledge, 9(4), 100590. What this study did and found
- Lane, Williams and Broecke (2023). The impact of AI on the workplace: Main findings from the OECD AI surveys of employers and workers. OECD Social, Employment and Migration Working Papers, No. 288. What this study did and found
- Reif, Larrick and Soll (2025). Evidence of a social evaluation penalty for using AI. Proceedings of the National Academy of Sciences, 122(19), e2426766122. What this study did and found
- Edmondson and Bransby (2023). Psychological Safety Comes of Age: Observed Themes in an Established Literature. Annual Review of Organizational Psychology and Organizational Behavior. doi.org
- Büchsenschuss, Koch-Bayram, Biemann and Puranam (2026). The impact of generative AI adoption on organizational networks: evidence from a field experiment. INSEAD Working Paper 2026/01/STR. Not peer-reviewed. sites.insead.edu
Related practices
- Measure readiness and resistance separately Staff can be keen on AI and worried about it at once. Ask about the two separately. A single “readiness” score hides the worry. Do next · Effort: medium (weeks, a few people)
- Budget for people and process, not just tools Licences are the small part of the cost. Plan most of the time and money for training, redesigning work, and helping people through the change. Do first · Effort: low (days, one person)
- Keep the entry-level tasks that newcomers learn on Before you automate routine work, check whether it is how juniors learn the job. If it is, keep some of it for them, or replace the learning another way. Do later · Effort: medium (weeks, a few people)