Trust, rules and responsible use
People adopt AI faster when the rules are clear and they feel safe to learn. This module covers policy, trust and fairness.
Learning objectives
- Explain how clear rules reduce shadow AI
- Draft a staff acceptable-use policy
- Build psychological safety for experimentation
- Apply fairness and transparency to AI that affects employees
5.1 Clear rules make people braver
Uncertainty about what's allowed slows adoption and increases risk: careful people don't use AI, while others use unapproved tools with sensitive data (shadow AI). A short, clear acceptable-use policy plus approved tools solves both.
5.2 What a staff AI policy covers
- Approved tools — which ones, for what, and how to request new ones
- Data rules — what can and can't go into AI tools (personal, confidential, client data)
- Checking output — people are responsible for what they use; verify facts, figures and sources
- Transparency — when to disclose that AI was used
- Prohibited uses — for example, making decisions about people without review
- Help and reporting — who to ask, how to report a problem
Keep it to a page or two, in plain language, with examples. Update it as tools and rules change. See AI Governance for the wider framework.
5.3 Psychological safety
People learn new tools by experimenting — which means making mistakes. If errors are punished or ridiculed, people stop trying or hide their use. Leaders build safety by sharing their own failed attempts, thanking people who raise problems, and treating early mistakes (within the rules) as learning.
5.4 Being fair and transparent with employees
- Tell staff when AI is used in decisions that affect them, such as hiring, performance or scheduling
- Be cautious about AI monitoring of employees — it damages trust and may breach data protection and employment law
- Remember that AI used for recruitment, promotion, termination, task allocation or performance monitoring is classed as high-risk under the EU AI Act
- Check impacts across groups — for example, whether training and tool access reach part-time, older or frontline staff equally
5.5 Wellbeing and workload
AI can remove drudgery — or raise expectations so that people are simply asked to do more, faster. Watch for rising workload, constant context-switching and anxiety. Ask people how the change feels, not just whether metrics improved.
Draft a one-page acceptable-use policy for your team using the six headings in 5.2. Include at least two concrete examples of what's allowed and what isn't.
See a worked example
Example lines: Approved: the company AI assistant for drafting, summarising and research. Data: never paste customer personal data or unreleased financial results into any tool other than the approved assistant; never paste them into free consumer tools. Checking: you are responsible for anything you send — verify facts, figures and quotes. Disclosure: tell clients when a deliverable was substantially AI-drafted. Prohibited: using AI to score or rank colleagues. Help: ask in #ai-help; report incidents to the AI team.
Key takeaways
- Clear rules plus approved tools reduce risk and speed adoption.
- A good policy is short, plain and full of examples.
- People learn by experimenting — make mistakes safe.
- AI in employment decisions is high-risk under the EU AI Act.
- Watch workload and wellbeing, not just metrics.