AI Changes Company Culture Through the Conversations Leaders Have

By Kristin Schleicher

The relationship between AI and company culture becomes tangible in everyday conversations: what employees feel safe asking, how leaders answer, and whether those answers lead to action. When a company introduces AI, employees are being asked to learn more than a new tool. They are also trying to understand what it means for their expertise, their responsibilities, and their future.

Will the time saved make room for better work—or simply more work? Can they question an AI-generated recommendation? Will asking for help be treated as learning or as falling behind?

These are reasonable questions. How leaders respond tells employees something about the organization they belong to.

My view is that leadership communication is central to AI adoption: making it possible to ask difficult questions, giving meaningful answers, and following through. An invitation to speak means little if the conversation goes nowhere.

Productivity does not tell the whole story

AI can improve performance, but the benefits depend on the task and the people using it. Research by Erik Brynjolfsson, Danielle Li, and Lindsey Raymond found that an AI assistant improved productivity in customer support, with larger benefits for less-experienced workers. Experienced workers saw smaller gains, and some quality measures declined. Those findings describe a particular workplace setting, rather than a universal outcome. Generative AI at Work

For HR and L&D, this raises an important question: what support does each group need?

A new employee may need help evaluating a suggested answer. An experienced colleague may need space to challenge it and explain what the system misses. Giving everyone the same introductory training will not necessarily meet either need.

“Any questions?” is only the beginning

During an automotive change project at Capgemini Invent, I worked with colleagues on employee feedback and change communication. Our analysis identified gaps between management communication and employee understanding.

That experience informs how I think about AI adoption. Announcing the purpose of a change does not establish that employees understand its implications.

A message about efficiency might leave people wondering whether their role is at risk. An invitation to experiment might feel uncomfortable if mistakes are penalized.

And “We welcome your questions” can become frustrating when every difficult question receives another reassuring phrase.

Leaders do not need to pretend they have every answer. They do need to explain what is known, what remains undecided, who is responsible, and when employees will hear more.

“We haven’t decided how the saved time will be used. We’ll review the pilot with the team before setting new targets, and we’ll update you on Friday” gives people something concrete—provided that review and update actually happen.

Three ways to make the conversation useful

1. Ask while the answers can still influence the plan

Start with a clearly defined task, such as drafting routine customer responses. Ask the people doing that work where assistance would help, what could go wrong, and what needs human judgment.

Agree on how the pilot will be evaluated. Look at quality, correction time, and workload alongside speed. Invite employees to identify problems without treating every concern as resistance.

Then explain which suggestions will be acted on, which will not, and why. Participation does not require agreement on everything. It does require a response.

2. Make learning expectations clear—and achievable

Provide practice using realistic tasks, clear guidance on permitted data, and time to check outputs. Teach people when to use AI, when to question it, and when to seek help.

Offer written examples, demonstrations, and opportunities to practice with colleagues. This can make learning more accessible across language backgrounds, experience levels, and learning preferences.

If managers say learning matters, they need to explain where it fits into the workload—and make room for it.

3. Explain who remains accountable

Employees should know which decisions AI supports, who reviews its output, and who is responsible when something goes wrong. They also need a clear route to question a recommendation, particularly when it affects their work or opportunities.

NIST’s voluntary AI Risk Management Framework offers a useful foundation for considering trustworthiness throughout the design, use, and evaluation of AI systems. Organizations still need to translate that guidance into responsibilities and everyday practices. NIST AI Risk Management Framework

“Human oversight” becomes meaningful when people know whose oversight it is.

AI and company culture: Communication needs a feedback loop

A shared sense of “we” becomes more credible when employees can see that their knowledge matters and their questions receive attention.

Leadership communication earns credibility when experience matches the answer: promised learning time appears in the schedule, concerns reach someone who can address them, and updates arrive when expected.

Before expanding an AI initiative, ask employees:

What has become easier, what has become harder, and what still feels unclear?

Then come back with answers—and show them what their feedback changed.