A demonstration is designed to show what a system can do. A public service, classroom or workplace needs to know something more difficult: what happens when the system gets it wrong. That shift changes the conversation from impressive output to responsibility.
Ask who owns the decision
UNESCO’s Recommendation on the Ethics of Artificial Intelligence places human rights, transparency and human oversight at the centre of its approach. Those principles become concrete when an organisation can name the person responsible for a decision assisted by a machine.
If an answer changes a person’s access to an opportunity, the review process should be easier to find than the sales presentation. Who can explain the result? What evidence can the affected person provide? Can the decision be reconsidered?
Test the awkward example
An average performance score does not explain every use. Before relying on an AI tool, an organisation can test how it handles incomplete records, ambiguous language and questions outside its intended purpose.
A useful trial records failures alongside successes. The point is to discover where a person needs to intervene and whether staff have enough time and authority to do so. A human approval button means little if nobody can realistically check the work.
Make limits visible
Readers should distinguish a tool that helps draft a document from a system that decides something about a person. Those uses involve different consequences and deserve different levels of scrutiny.
The strongest explanation is often plain: what the system contributes, what it cannot reliably establish and what a person must verify. Confidence should come from a process that can be examined, rather than from the fluency of an answer.
