GLOSSARY
Explainability
Explainability means being able to provide understandable reasons for an AI system's results or behaviour; a plausible self-explanation alone does not establish its actual workings.
A recommendation must be open to challenge
If AI rates a customer as an unlikely prospect, salespeople need a basis they can examine. Is the reason a lack of need, an incorrect product match or simply missing data? These explanations imply different next steps. A generic label such as “poor fit” provides little help. Explainability becomes useful when it exposes relevant information, decisive criteria and important limitations in a form that allows a knowledgeable person to examine and, where appropriate, correct the recommendation.
A plausible explanation may not reveal the actual cause
A convincing explanation does not establish how an output was produced. Different methods support different forms of explanation. The application should specify what kind of traceability is needed and how it will be tested. Evidence and recorded inputs assist professional review. A test can deliberately change relevant information and examine whether the recommendation and explanation respond appropriately. Human involvement becomes more effective when reviewers also have time and authority to intervene. Explanations must remain understandable without implying a precision the system cannot substantiate.