Responsible AI Stewardship — Protecting People While Adopting AI

Artificial intelligence is moving quickly into everyday work. It can help us draft, analyse, create, and make sense of complexity at remarkable speed. Yet responsible AI is not only a technical governance issue. It is also a leadership, culture, and psychological safety challenge.

This article reflects on a recent talk for the Coaching Ethics Forum on responsible AI stewardship. As adoption accelerates, leaders need to pause and ask a more human question: How do we protect people, trust, and accountability while adopting AI?

Beyond a technology decision

Responsible AI stewardship begins when an organisation asks not only, “How do we adopt AI?” but also, “What kind of workplace are we creating as we adopt it?”

Too often, AI is treated as a technology decision: selecting tools, setting policies, managing data, and measuring productivity. These are important considerations. But AI adoption is also a cultural and leadership decision. It shapes how people think, learn, collaborate, make decisions, and relate to their own expertise.

The KPMG and University of Melbourne global study found that 58% of employees use AI regularly at work. Yet responsible-use guidance, governance, and AI literacy have not kept pace. Many employees reported non-disclosure, policy breaches, and reliance on AI-generated outputs without sufficient evaluation (Gillespie et al., 2025).

For leaders, this creates a tension. We want people to experiment and innovate, but we also need to protect sound judgment, privacy, quality, and trust. The question is not whether AI will change work. It already is. The question is whether we will actively shape that change, or allow it to shape us.

The hidden AI workplace

AI does not simply automate work away. It can also intensify work.

Research by Ranganathan and Ye (2026) suggests that AI can expand the scope of work, blur the boundaries between work and non-work, and increase multitasking. What initially feels like greater momentum can quietly become workload creep, cognitive strain, and less time for recovery.

Alongside this sits the hidden AI workplace: a reality in which people use AI, benefit from it, and may even depend on it, while feeling unable or unwilling to speak openly about how they use itThe KPMG and University of Melbourne study found that more than half of workers avoid revealing when they use AI, and many present AI-generated material as their own (Gillespie et al., 2025).  This is not merely a compliance issue. It is a cultural signal.

When AI use is hidden, leaders lose visibility into how work is being done, where risks are emerging, and what useful practices could be shared. Teams also lose an opportunity to learn together. Hidden use can weaken accountability, distort performance expectations, and make it harder to distinguish thoughtful experimentation from harmful drift.

Why people stay silent

People do not usually hide AI use because they are resistant to innovation. Often, they are making a rational calculation about the personal cost of disclosure.

Employees may worry that revealing an effective AI workflow will result in more work rather than better work. They may fear being seen as less capable, losing a competitive advantage, breaching unclear rules, or making their role easier to replace. Anicich and Brouwers (2026) found that organisational trust and psychological safety are powerful predictors of whether employees share or withhold AI knowledge. Employees with lower levels of trust and psychological safety were far more likely to keep AI-related workflows, prompts, and techniques to themselves.

This is where the psychological challenge begins.

AI can quietly reshape how people think, rely on support, and disclose their work. Cognitive offloading may make difficult tasks feel easier, but it can also weaken the habit of independent thinking. AI support may be useful, but when does support become dependency? Openness may be ethical, but what happens when disclosure is met with stigma rather than curiosity?

These questions matter because they sit beneath the surface of AI adoption. A workplace can have a policy, an approved tool, and a governance framework — while still lacking the psychological safety people need to disclose, question, challenge, and learn.

Ethical drift happens quietly

Ethical drift rarely begins with bad intent.

It often starts with a small compromise: an urgent deadline, a pressure to reduce cost, a desire to keep up with competitors, or an assumption that “just this once” will not matter. Gradually, the exception becomes normal. A lower standard becomes the new baseline.

AI can accelerate this pattern. Faster outputs and easier access to information can create an illusion that more speed always equals more value. Yet speed without reflection can lead to poor-quality decisions, overlooked bias, weak accountability, and an erosion of human judgment.

The mirror question for leaders is simple: Are we using AI to strengthen our capacity for wise judgment, or are we allowing convenience to replace it?

Responsible AI stewardship requires moral attentiveness. Leaders need to notice not only whether AI works, but what repeated AI use is doing to standards, reflection, accountability, and the lived experience of work.

Stewardship means protecting while adopting

Responsible AI stewardship is not about slowing innovation. It is about holding innovation with responsibility.

Stewardship means looking beyond immediate efficiency gains and asking what is sustainable for people, teams, customers, and society. It means recognising that leadership is not simply about introducing a tool; it is about creating the conditions in which that tool is used with dignity, discernment, and accountability.

A practical framework is to see, speak, safeguard, and steward:

•   See: Surface risks before they cascade. Notice hidden AI practices, changing workloads, emerging dependencies, and the small compromises that may lead to ethical drift.

•    Speak: Make AI use discussable rather than shameful. Create space for people to ask questions, disclose experimentation, share learning, and raise concerns without fear.

•     Safeguard: Protect autonomy, dignity, data, and human judgment. Be clear about where AI is appropriate, where it is not, and where human review is essential.

•     Steward: Keep accountability human. Ensure that responsibility does not disappear into a system, process, or platform.

The distinction here is important: governance sets the guardrails; stewardship asks leaders to remain awake at the wheel.

What leaders can do now

Leaders do not need perfect certainty before acting. They do, however, need clear intent, practical boundaries, and the courage to model responsible use.

•  Create explicit norms for AI disclosure, acceptable use, and human review, so employees are not left navigating ambiguity alone.

•  Build psychological safety around AI experimentation, so sharing a useful workflow is recognised as a contribution rather than treated as corner cutting.

•  Protect boundaries through intentional pauses, more thoughtful sequencing of work, and space for reflection and human connection.

•  Invest in AI literacy across the organisation, not only in technical teams, so that judgment and governance are shared rather than concentrated.

• Clarify where accountability sits, especially for higher-risk decisions, because stewardship fails when responsibility becomes diffuse.

•  Decide openly how AI-generated efficiency gains will be used: for deeper analysis, innovation, development, or recovery — not simply to add more work.

The future of AI at work will not be shaped only by the sophistication of our tools. It will be shaped by the courage of our cultures, the quality of our questions, and the standards we refuse to let drift.

Responsible AI stewardship invites leaders to adopt AI with curiosity and ambition, while keeping people, judgment, and human flourishing at the centre of the journey

Reference:

Anicich, E.M. and Brouwers, J. (2026) ‘Why employees aren’t transparent about their AI usage’, Harvard Business Review, 10 June. 

Gillespie, N., Lockey, S., Ward, T., Macdade, A. and Hassed, G. (2025) Trust, attitudes and use of artificial intelligence: A global study 2025. Melbourne: The University of Melbourne and KPMG. DOI: 10.26188/28822919.[1]

Ranganathan, A. and Ye, X.M. (2026) ‘AI doesn’t reduce work—it intensifies it’, Harvard Business Review, 9 February.