A company announces that AI should help people experiment. In the next project meeting, the fastest draft receives enthusiastic praise. The colleague who asks for time to check its assumptions is told to keep things moving.
Which message should employees believe?
AI and company culture meet in these ordinary moments. The policy explains what the organisation intends. Repeated reactions help employees interpret what earns approval, what threatens their professional standing and which kinds of effort are worth showing.
For leaders, the useful question is therefore specific: when someone uses AI, questions an output or asks for help, what does the team learn from the response?
Make culture observable
“Culture” can become an explanation that ends a conversation. A deployment stalls, so the organisation concludes that people have the wrong mindset.
That diagnosis is difficult to test. It may also obscure a reasonable response to unclear expectations or unsuitable work.
Here, culture means the shared expectations people develop about how work is done and how others will respond. This is a practical working definition, not a validated assessment model.
Look for situations colleagues can describe. Who receives credit for a completed proposal? Is asking a basic question treated as responsible preparation or a sign of incompetence? Does careful review count as contribution? What happens when an experienced employee changes their mind?
These questions connect culture with behaviour that can be examined. They also leave room for different explanations. A rushed response may reflect an exceptional deadline rather than a settled norm. Establish a pattern before labelling a whole team.
What current research can tell us
Three contemporary sources make this conversation relevant without settling how a particular company should act.
Gallup's August 2026 analysis of workplace culture draws on its first-quarter US workforce study. Employees in AI-implementing workplaces reported both positive and negative changes in culture. Perceived manager support was associated with more favourable accounts. These are employee perceptions and associations, not evidence that AI or a manager's behaviour caused the change. The article also discusses a separate CHRO survey; that group should not be confused with the employee sample.
Deloitte's UK GenAI Workforce Survey 2026, with fieldwork in May and June, reports that some workers perceive a stigma around workplace AI use. The wider survey covered 25,000 UK working adults aged 18–70, with different bases for individual questions. A reported stigma warrants investigation; it does not establish that every workplace penalises AI users.
Microsoft's May 2026 Work Trend Index highlights a perceived mismatch between ambitions for AI-enabled change and workplace incentives. Its main survey covered 20,000 AI-using knowledge workers in ten markets, conducted from 18 February to 7 April. It excluded non-users. The survey findings do not prove that changing rewards will cause better outcomes.
These commercial institutions offer useful primary evidence, with different populations and interests. The practical analysis that follows is an editorial proposal, not a tested intervention derived from those surveys.
Follow a proposal through an ordinary meeting
Imagine a fictional commercial team at a wholesaler preparing a seasonal range proposal.
An employee uses an approved AI assistant to organise suitable product material into a draft. A colleague notices that the wording makes a delivery assumption the underlying information does not support. Another colleague helps compare the draft with the agreed brief.
At the review meeting, the manager praises how quickly the document appeared. The correction is acknowledged briefly, then the discussion turns to why the proposal is not ready to send.
Nobody explicitly says that checking is unimportant. Yet the visible distribution of praise could suggest that producing the draft is valuable and reviewing it is a delay.
If that response keeps recurring, employees have a reason to prioritise the part of the work that gets recognised. The organisation may believe it rewards responsible AI use while making one essential part of that responsibility hard to demonstrate.
This is an invented example, not an observed deployment or a Learning Coach client case. It illustrates a possible contradiction, not an inevitable employee reaction.
Ask what now counts as professional competence
AI use can complicate the meaning of a familiar accomplishment.
If a strong first draft once signalled expertise, colleagues may be unsure how to interpret one produced with assistance. Some may dismiss legitimate work as effortless. Others may treat a polished output as sufficient proof that its author understands the subject.
Neither judgment examines the contribution.
In the fictional team, the relevant questions concern how the employee selected the material, identified missing information, adapted the proposal and established what could responsibly be sent. The assistant's involvement is part of that explanation.
Managers can make these contributions visible by discussing the work rather than guessing how much effort it should have taken. Ask which decision required judgment and what changed during review.
The colleague who identified the unsupported delivery assumption contributed something specific. Naming that contribution gives the team a clearer account of competence than either automatic praise for AI use or automatic suspicion of it.
Examine the response to uncertainty
A statement such as “we welcome questions” becomes meaningful when someone asks an inconvenient one.
Suppose an employee admits that they do not understand why a generated recommendation fits the proposal. The team could respond with embarrassment, impatience or a practical examination of the reasoning.
That response can affect what the employee feels able to bring forward next time. It does not follow that every question requires a lengthy group discussion. A manager might arrange a short follow-up with the right colleague while keeping the meeting moving.
What matters is that the uncertainty has a credible destination.
Leaders should be accurate about the conditions of such conversations. They should not promise that every mistake will have no consequences or that every concern can be kept confidential. They can explain what needs investigation, who will help and what the person should do while the question remains open.
An invitation to speak loses credibility if the next response leaves the employee exposed and the issue unresolved.
Give review and shared learning visible credit
Recognition is broader than a formal reward scheme. It includes whose work is mentioned in a meeting, whose help is remembered and what a manager asks about.
If only the person presenting an AI-assisted output receives credit, the people checking, explaining and improving it may disappear from the account of success.
For the fictional proposal, a manager could distinguish the useful initial structure from the later correction and clarification. That provides a more accurate account of how the final document became usable.
The same principle applies to someone who helps a colleague understand a recurring mistake. If teaching others is expected, it needs to be recognised as work and given appropriate time. Otherwise, an employee may face a choice between meeting their own delivery expectations and providing the support the organisation says it values.
This is not a recommendation to reward every visible AI activity. Recognition should relate to a useful contribution. Producing more prompts or appearing enthusiastic does not establish that contribution.
The broader human learning advantage depends partly on capabilities colleagues can share, explain and apply beyond the first assisted answer.
Preserve the habits that still serve a purpose
The phrase “old habits” can imply that established practice is automatically an obstacle. Some habits protect quality, preserve context or make coordination reliable.
In the wholesale example, checking a delivery statement against an authoritative source remains useful whether a person or an assistant drafted the sentence. An established review conversation may contain knowledge missing from the written template.
Before replacing a habit, ask what function it serves.
Perhaps the check is necessary but the handoff is cumbersome. Perhaps several people repeat the same review because responsibility is unclear. Perhaps a familiar step has outlived its purpose.
Each possibility calls for a different change. Respecting the function of an existing practice does not require preserving its exact form. Equally, introducing AI does not justify discarding it without understanding what would be lost.
Employees' objections can help reveal that function. Assess the explanation rather than treating either experience or novelty as decisive.
Allow for different local experiences
A company can have a shared AI policy and several different everyday cultures.
One team may treat assistance as ordinary, another as something to hide. A junior employee may see asking for help differently from a senior colleague with an established reputation. Remote staff may miss the informal explanations available to people sitting together.
These are possibilities to investigate, not assumptions about particular ages, nationalities or job titles.
Ask people to describe recent situations. What happened when someone shared an assisted draft? Was a correction welcomed? Did the person doing the review have enough time? Compare accounts with the work and its constraints.
Avoid relying only on the most confident advocates or the most memorable complaint. A local discussion should make room for people whose experience does not fit the dominant story.
The aim is to identify which expectations are actually shared and where interpretations diverge.
Change a recurring interaction, then examine it
A culture programme does not have to begin with a new slogan.
The fictional team could start with its proposal review. The manager could explain that a useful draft, a necessary correction and a clear account of uncertainty are all legitimate contributions. The meeting could examine whether the document is fit for its purpose, with credit reflecting the work involved.
That change also needs practical support. If reviewing is expected, it must fit the schedule. If an employee identifies missing information, someone needs responsibility for obtaining it. Praise cannot compensate for an impossible workload.
After several relevant reviews, ask what changed. Are colleagues identifying uncertainties earlier? Is useful review work more visible? Are discussions becoming longer without resolving anything? Do employees still interpret the standard differently?
This is a proposed local inquiry, not a guarantee or a validated culture score. An unsuccessful change should be revised rather than reported as transformation because it was announced.
Bring the ordinary moment into the leadership room
For a corporate event, the fictional proposal meeting offers a concrete discussion.
Ask participants to describe what each colleague contributed and what the manager's response appeared to value. Then ask them to write a response that recognises quality without overlooking delivery obligations.
Different interpretations can reveal why a general statement about responsible AI use may leave employees uncertain.
A keynote can frame the question and challenge easy labels such as “enthusiasts” and “resisters.” A facilitated session can examine examples and identify a recurring interaction worth changing. Our guide to designing a corporate AI learning summit explores how to connect that discussion with workplace follow-through.
An event cannot establish what employees experience across an entire organisation. That requires subsequent listening, observation and action.
Make the everyday standard credible
Sources were checked on 28 October 2026. This article uses selected public survey analyses, not a systematic review or causal evaluation. Deloitte's page does not display an exact publication date; its stated fieldwork period is identified above. Gallup's culture article does not provide the full employee sampling detail on that page.
Learning Coach's practical argument is that people need a credible account of good work as AI changes how it is produced. That account includes the visible result and the less visible contribution that makes the result dependable.
Before asking employees to change their habits, examine a recent reaction of your own: what did it teach the team to value?
Bring that question to your next leadership or workplace learning event. Explore Learning Coach's international keynotes and request a tailored proposal.
References
Meinen, M., and Mulherin, M. AI's Effect on Workplace Culture. Gallup, 16 August 2026. US employee perceptions and a separate CHRO survey.
Deloitte UK. Transform your workforce and your organisation will follow. GenAI Workforce Survey 2026, May–June fieldwork. Public thematic analysis accessed 28 October 2026.
Microsoft. Agents, human agency, and the opportunity for every organization. Work Trend Index, 5 May 2026. Survey methodology and population limits consulted.
