Trust, Ethics and Continual Improvement with AI
Domain focus
Trust, ethics, and continual improvement with AI covers the evidence and behaviors that make AI acceptable in service management, and how to improve it over time. The exam tests ethical principles, trust signals, and the use of continual improvement rather than one-off approval.
Core rules
- Trust is multidimensional: transparency, explainability, fairness, privacy, security, reliability, accountability, and human agency. Accuracy alone is not trust.
- Make ethics operational: identify affected stakeholders, assess harms and benefits, provide a route to question or appeal a decision, and record who is accountable.
- Apply the continual improvement model: what is the vision, where are we now, where do we want to be, how do we get there, take action, did we get there, and how do we keep momentum.
- Monitor drift, bias, incidents, overrides, complaints, and user confidence. Set review triggers for model changes, new data, or new use.
- Communicate limitations in plain language and publish what is automated, what is reviewed, and how to escalate.
Worked example
An AI tool summarizes customer emails for agents. Agents report missing complaint details. Reasoning: the outcome is not just summary accuracy; it is risk of missed obligations. Add a transparency note, require human review for complaint and vulnerability keywords, measure omission rate against a sample, add an appeal/correction route, and review after model changes. If omission persists, narrow the use case or pause it. Trust improves through evidence and remedy, not a slogan.
Traps
- Equating trust with high accuracy scores.
- Running a one-time ethics review and never revisiting it.
- Hiding automation from users or removing human escalation.
- Ignoring complaints and overrides as improvement data.
- Treating continual improvement as a yearly audit only.