Building trust when AI changes how decisions are made
As AI starts shaping more workplace decisions, trust depends less on the tool itself and more on how leaders explain, govern and challenge its use. Research suggests people respond better when leaders keep human judgement visible, ask better questions, and set clear guardrails.
AI is changing the path to decisions, not just the speed of them. That shift can unsettle people because it raises a quiet question: if a system drafts, ranks or recommends, who is really responsible? Trust rarely disappears because of one model or one poor output; it erodes when decisions feel opaque, inconsistent or unchallengeable. Evidence from Stanford, Harvard and MIT points to a practical lesson for leaders: confidence grows when AI is introduced with clear purpose, visible human oversight and honest limits 123. The aim is not to slow useful change. It is to make sure people can understand, question and stand behind the decisions being made.
Make the decision path visible
The first trust test is transparency about where AI sits in the decision chain. Stanford’s AI Index shows that adoption is broad and capability is moving quickly, while responsible AI practices are not keeping pace and incidents continue to rise 1. For leaders, that means people need more than reassurance that a tool is being used. They need to know what the system does, what it does not do, and where a person must still review or override it.
A practical approach is to map each important decision and identify the steps that are automated, assisted or fully human. Then explain those boundaries in plain language to managers and teams. If AI is used to sort applications, prioritise work, draft performance notes or flag risks, say so. Silence invites suspicion; clarity helps people understand that the organisation is using AI as an input, not a substitute for accountability.
Keep human judgement in the room
Harvard Business School researchers found that leaders who performed well with AI agents also tended to perform well with human teams, and that stronger leaders asked more questions, encouraged turn-taking and used more collaborative language 2. That finding matters because it suggests trust is strengthened not by acting as though AI knows best, but by leading in ways that draw out better judgement from people around the table.
In practice, this means AI should support discussion rather than end it. Use it to surface options, test assumptions or widen the search for evidence, then ask people to challenge the output. A good meeting question is not ‘What does the model say?’ but ‘What might it be missing, and who can test that?’ When leaders show that AI is one source of input among several, they reduce the sense that decisions are being handed down by a black box.
Choose governance that people can understand
MIT Sloan’s work on generative AI highlights a useful distinction between broadly applicable tools and tailored solutions, each bringing different risks, governance needs and cost pressures 3. That distinction is helpful because many trust problems arise when organisations manage all AI as though it were the same thing. A general productivity tool used by individuals raises different issues from a model embedded in a core workflow.
Leaders should therefore set separate rules for different uses. For everyday tools, establish clear guidance on acceptable prompts, sensitive data and review expectations. For tailored solutions, define who approves the use case, who monitors outputs, and how exceptions are handled. Keep the governance simple enough that managers can apply it without guesswork. If policy is too technical, people will either ignore it or follow it ritualistically without understanding the risk.
Build confidence through explanation and challenge
Trust is not created by claiming AI is accurate all the time. It is created when leaders are honest about uncertainty and create space for challenge. Stanford’s research notes the gap between capability and responsible deployment, which is a reminder that even impressive systems can behave unevenly depending on the task 1. That makes explanation essential. When an AI recommendation is accepted, managers should be able to say why; when it is rejected, they should be able to explain that too.
An understated Authenticate advisory perspective is that leaders often underestimate how much confidence comes from process rather than polish. If people can see the reasoning, the checkpoints and the escalation route, they are more likely to trust the decision even when they do not fully trust the system. A short record of how a recommendation was reviewed can be more valuable than a long policy nobody reads.
Support people through the change, not around it
MIT Sloan also notes that trust, collaboration and open-mindedness are essential as organisations introduce AI into people management and wider work processes 4. That is an important reminder that adoption is emotional as well as technical. People may worry about status, fairness or being judged by systems they do not control. Leaders do not need to promise that those concerns are unfounded. They need to address them directly.
Practical steps include involving representatives from different functions early, using pilot groups to test whether the workflow feels fair, and training managers to explain AI-supported decisions in ordinary language. Also make room for feedback from the people affected. If staff can raise concerns without penalty, leaders are more likely to spot where a process is confusing or feels misaligned with organisational values. Trust grows when people see that AI is being introduced with them, not to them.
Sources
- The 2026 AI Index ReportStanford University — Human-Centered AI · University research · link checked 15 October 2026
- Think You’re a Good Leader? Try Leading an AI TeamHarvard Business School AI Institute · University research · link checked 15 October 2026
- How to manage two types of generative artificial intelligenceMIT Sloan School of Management · University research · link checked 15 October 2026
- Who should be the head of generative AI — and what they should doMIT Sloan School of Management · University research · link checked 15 October 2026
Further professional reading
External professional resources, separate from the numbered sources above. These are not university studies, and listing them does not indicate any collaboration, certification, affiliation or endorsement.
- Global Institute of Directors (opens in a new window)External professional resource · governance and directorship
- World Coaching Institute (opens in a new window)External professional resource · coaching practice