Why Does a Team Stall on AI When Nobody Knows Who Answers for a Mistake?

The AI tool was approved, paid for, and announced on a Monday. Three weeks later, almost nobody had used it. Nobody had objected either.
Based on my professional experience and opinion, not clinical fact.
I am Sahar Andrade, MB.BCh, a Neuroleadership Coach and Forbes Coaches Council member, and here is the direct answer. A team can stall on AI when nobody knows who answers for a mistake, because stopping can feel like the safest move, and the question of who owns the output can go unasked. A leader can loosen the stall by saying, before the work starts, who checks, who decides, and who handles a problem.
What Does a Stalled AI Rollout Sound Like?
Here is an illustration, not a case. A leader rolls out an AI drafting tool. There is a demo and a short training. Three weeks later, usage is flat. The leader books another training and sends a sharper reminder. Nothing moves.
Then a senior analyst asks one question in a one-to-one. "If the tool gets a number wrong in a client report, is that on me?" The leader starts to answer and stops. Nobody had ever said.
On a team, the stall rarely sounds like fear. It sounds like "I am still getting used to it" and "I will try it on the next project." People use the tool for a meeting agenda, then go back to the old way for anything with a client's name on it.
What Do Leaders Get Wrong About a Team That Will Not Use AI?
The common reading is resistance to change, or a skills gap. So the fix becomes more training and more pressure.
This is not a training gap. It is an ownership gap.
Look at it from the team's chair. If I use the tool and it is right, nobody notices. If I use it and it is wrong, my name is on the report. If I do not use it, nothing happens to me. From there, stopping is the safest move on the board.
Teams are also hearing two AI stories at once. One, told to investors, is about efficiency. The other, told in the town hall, is about helping people do better work. When the stories do not match, people fill the gap themselves.
What Does the Brain Do With an Unclear Risk?
The brain can read uncertainty about consequences as a kind of threat. A social one: will I be blamed, will this cost me something I spent years building? Under threat, attention can narrow toward whatever carries the least exposure. In most workplaces, the least exposed move is to do nothing new.
In 1999, Amy Edmondson studied 51 work teams inside one manufacturing company. Teams where people felt safe to take interpersonal risks did more learning behavior: they asked for help, raised problems, and talked about errors.
One honest limit: that was one company in the 1990s, and it did not study AI. In that study, psychological safety was associated with more team learning behavior. In my view, it points to something practical: what people expect will happen to them shapes whether they speak.
Is It Not Obvious That the Person Using AI Owns the Output?
It is obvious to the leader. It has not been said. Under pressure, people act on what they have been told and on what they saw happen to the last person who made a mistake.
And look at the rule itself. If a person owns every line the tool produces and has not been given the time to check every line, they have not been given ownership. They have been given exposure. Ownership works when it comes with a boundary.
What Can a Leader Say Before AI Work Starts?
Three sentences, out loud, for any piece of work with AI in the middle.
Who checks. "The AI drafts, and Maria checks the numbers against the source file." One name, and a clear line around what is being checked.
Who decides. "I make the final call on what goes to the client." The judgment at the end sits with a person.
Who handles it. "If something slips through, bring it to me the same day. We fix it together, and I talk to the client." This is the sentence most leaders skip, and it is the one the analyst was asking for.
Two things help it hold. Ask the tool to show where its answer came from, so the person checking has something to check against. And let the first mistake be the test: if it becomes a review of the process, people keep using the tool and keep reporting what they find.
You can't give what you don't have. A leader cannot give a team clarity about AI that the leader has not worked out first.
Listen to the full episode of AI Café Conversations: https://www.saharconsulting.com/ai-cafe-conversations-podcast
Want to see what a team reads in its leader before anyone says a word?
Get What Your Team Can Tell:
Frequently Asked Questions
Why does a team stall on AI when nobody knows who answers for a mistake?
Because, from the team's side, not using the tool can be the safest choice. If the output is right, nobody notices. If it is wrong, a name is on it. When nobody has said who checks the work, who makes the final call, and who handles an error, people can quietly go back to the old way. The stall can ease once a leader answers those three questions out loud before the work begins.
Is low AI adoption a training problem?
Sometimes, and that is worth ruling out first. When a team has been trained, can use the tool, and still avoids it for important work, more training rarely moves anything. In my view the more useful question is about ownership: does each person know what they are responsible for checking and what happens if something slips through? A team that can answer that is far more willing to use what it was trained on.
Who is responsible when AI makes a mistake at work?
That is for each organization to decide and state clearly, and this piece is not legal advice. As a leadership practice, three things should be named before the work starts: one person who checks the output and what exactly they check, one person who makes the final call, and one person who steps in when something goes wrong. When those are named, people stop guessing, and guessing is what slows a team down.
What should a leader say when someone asks "is this on me?"
Pause before answering. A rushed "of course not" or a rushed "obviously, yes" both close the conversation. Then answer with specifics: what the person is responsible for checking, what they are not, who decides, and who they bring a problem to. If the honest answer is "we have not decided yet," say that, and give a date by which it will be decided. People can work with a clear not-yet.
How should a leader respond to the first AI mistake on a team?
Handle it as a review of the process, in front of the team. What was checked, what was missed, and what changes next time. The first error teaches everyone what the rules were really worth. If it becomes a search for the person, people learn to stop using the tool or to stop reporting what they find. If it becomes a fix, reporting stays open, and that is what keeps later errors small.
Does psychological safety matter for AI adoption?
Amy Edmondson's 1999 study of 51 teams in one company found that teams where people felt safe to take interpersonal risks did more asking, flagging, and discussing of errors. That research predates AI tools and did not study them, so it is not proof about AI adoption. In my view it points to something practical: what people expect will happen shapes whether they raise a problem, and AI work produces problems that need raising.
About the author
Sahar Andrade, MB.BCh, is a Neuroleadership Coach, Forbes Coaches Council member, and host of the AI Café Conversations podcast. A physician turned neuroleadership coach (non practitioner), she teaches executives how the nervous system shapes leadership under pressure.
Sahar Consulting LLC provides educational leadership development services. Our programs, frameworks, assessments, and content, including 4 Ka Leaders™ and B.R.A.I.N.™, are not medical, psychological, or mental health services, and they do not diagnose, treat, or cure any condition. References to the nervous system describe general leadership patterns, not clinical findings about any individual. Nothing here creates a physician-patient or therapist-client relationship. For health or mental health concerns, consult a licensed professional. In a crisis, contact a licensed mental health professional or your local emergency services.















Comments