We keep asking why employees won’t adopt AI. The right question is: why should they?
Walk through any organisation lately and you’ll likely find three groups of people. Small clusters who are rebuilding ways of working around AI and leveraging its capability. Larger groups who tried it a few times, got mediocre first results, and are drifting back to old habits. And a surprising number — including the experienced, senior, respected — who mostly dismiss it.
Your licence distribution and training completion dashboards might tell you a different story. Yet this is the reality at most companies and recent studies show it. According to McKinsey, 88% of organisations now use AI in at least one function, yet only 6% capture material bottom-line value. Deployment is nearly universal. Value remains elusive.
When the returns don’t come, most leadership teams will reach for levers they know well:
- more training
- more communication from the top
- increasing cost pressure — trimming teams until using AI is no longer optional.
In practice, while each can be an enabler of the change you want, none address the deeper structural causes of poor adoption.
Training builds necessary capability, but it falls short of impacting ways of working or organisational norms that are required for skilled AI usage to expand.
A mandate from the top does real work: it makes AI use officially legitimate, unlocks resources, and sets direction. Still you cannot mandate adoption from corporate headquarters, whether AI use is safe is decided and experienced team by team.
Forced automation can bring short term savings but it creates deep fear that drives concealment and resistance (and puts tremendous pressure on the already overstretched middle managers). It also destroys one of your organisation’s most valuable currencies: trust.
You cannot train or message people out of a social penalty, nor can you frighten them into trust.
The gap between what AI could do in your organisation and what people feel safe actually doing with it is where your value, engagement, and credibility quietly disappear.
The Unwritten Rules Your AI Strategy Ignores
People who use AI every evening on their own accounts don’t lack skills or interest — they lack workplace norms, evaluation practices and career signals that make using it openly safe and worthwhile. The human and organizational conditions for adoption.
And where adoption stalls, the cause is mostly the social and career cost of being seen to rely on AI — and the lack of human involvement in designing the change.
We believe that these cultural barriers are what you need to tackle, they fall into four categories:
- The visibility penalty: Being perceived as less competent or expert because of AI usage. Engineers reviewing identical code rated its author 9% less competent when told AI was involved, with harsher penalties for women and older workers (Acar et al., HBR, 2025). A PNAS study found the same stigma across every age, gender, and occupation.
- The identity threat: for professionals whose standing rests on deep expertise (radiologists, lawyers, and knowledge workers) AI is a challenge to a hard-won occupational identity. AI is a question mark over who they are professionally.
- The psychological safety risk: experimenting in public, disclosing AI use, admitting what you or your agent got wrong are all interpersonal risks in an organisation. People will only take these risks when being wrong or failing is explicitly survivable and safe in your culture.
- The team social dynamics: teams run on informal hierarchies, and if one person suddenly is twice as fast or proposes some work automation, it can signal a status claim and a threat to others— many in that case will choose concealment or conformity over innovation; to stay within the group accepted norms and avoid conflict.
Faced with the AI mandate, your people run a quiet risk–reward calculation: “Will using AI here improve my status — or make me look replaceable?” Until you can address these social risks, your AI roadmap will fail at the feet of your culture.
Coming soon in this series: “From unwritten rules to deliberate design — what you can do to redesign your cultural norms in support of AI adoption”.
Partner with Degree23
At Degree23, we help organisations navigate complex people transformations. We combine market insight and HR expertise with practical implementation experience to ensure your initiatives support both your people strategy and your business growth.
Let’s start the conversation. Contact us: hello@degreetwentythree.com
Interested in this topic? Check out these additional references:
McKinsey & Company, “The State of AI in 2025: Agents, Innovation, and Transformation,” November 2025 — https://www.mckinsey.com/capabilities/operations/our-insights/the-state-of-ai
MIT NANDA, “The GenAI Divide: State of AI in Business 2025,” July 2025 (preliminary report) — https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf
Acar, Gai, Tu & Hou, “Research: The Hidden Penalty of Using AI at Work,” Harvard Business Review, August 2025 — https://hbr.org/2025/08/research-the-hidden-penalty-of-using-ai-at-work
Reif, Larrick & Soll, “Evidence of a Social Evaluation Penalty for Using AI,” PNAS, 2025 — https://www.pnas.org/doi/10.1073/pnas.2426766122
Duke Fuqua Insights summary of the PNAS study, “Is AI Damaging Your Professional Image?”, 2025 — https://www.fuqua.duke.edu/duke-fuqua-insights/Is-AI-Damaging-Your-Professional-Image
Conceptual foundations: Edmondson, “Psychological Safety and Learning Behavior in Work Teams,” ASQ, 1999; Petriglieri, “Under Threat,” AMR, 2011 — https://journals.sagepub.com/doi/10.2307/2666999
