An AI feature can generate a response in seconds and still leave someone unsure what to do next. Is the information reliable? Can they change the result? What happens if they accept a recommendation that turns out to be wrong?
These questions connect usability to the wider responsibilities of building AI into digital products.
World Usability Day, observed on 12 November, brings that connection into focus. This year’s theme, “Designing a Sustainable Future with Ethical AI,” asks businesses and product teams to consider how their decisions affect the people using their technology.
For teams developing digital products, this means looking beyond whether an AI feature functions. It means examining whether people can understand it, use it appropriately and retain meaningful control.
Start with a problem worth solving
Before introducing AI, define the task it should improve.
A customer support assistant might help users find relevant information. A document tool might help employees prepare an initial draft. A recommendation feature might make a large catalogue easier to navigate.
Each use case needs a clear measure of success. Does the assistant help people resolve their questions? Does the drafting tool reduce the total time spent writing and checking? Do recommendations help customers find suitable products?
These questions give teams a basis for evaluating whether AI adds value. Counting generated responses or feature interactions alone cannot answer them.
Make the system’s role understandable
People need enough information to judge how to use an AI output.
For a drafting tool, that could mean clearly identifying generated content and explaining that it requires review. For a recommendation feature, it could mean explaining which preferences influenced the suggestions.
Transparency should help someone make a decision at the point where it matters. A lengthy explanation hidden in a settings menu may do little for a user deciding whether to accept an output.
NIST’s AI Risk Management Framework identifies transparency, accountability and explainability among the characteristics relevant to trustworthy AI. It also makes an important distinction: transparency alone does not establish that a system is accurate or fair.
Design for correction and control
Consider an AI tool that drafts a customer email. Generating the text is one part of the experience. The user also needs to review the recipient, check the claims, edit the wording and decide whether to send it.
The interface should support those decisions. Review controls, editable outputs and confirmation steps need to be designed alongside the AI feature.
The appropriate level of oversight depends on the use case and its consequences. NIST recognises that different human–AI arrangements require different approaches to oversight.
For product teams, the practical question is: where does a person need the ability to intervene, and does the design make that intervention possible?
Test understanding as well as completion
A user may complete a task while misunderstanding what the system has done.
During usability testing, teams should examine whether participants recognise AI-generated information, understand relevant limitations and know how to correct an unsuitable result.
Testing should also include situations where the feature performs poorly. What happens when a response is irrelevant, incomplete or unavailable? Can the user continue their task? Is there a clear route to support?
These scenarios reveal whether the product remains usable when the experience falls short of expectations.
Include sustainability in product decisions
The sustainability theme also invites teams to examine the resources their features require.
During planning, ask whether every interaction needs an AI-generated response, whether repeated processing can be avoided and whether a simpler solution could meet the same need.
Treat these as design questions to investigate and measure. A feature’s value should be assessed alongside the effort required to operate, maintain and use it.
Bring the right expertise into the process
Usability depends on decisions made across product design, development, project management and quality assurance. Clear requirements, understandable interfaces and thorough testing all contribute to the final experience.
Sphise connects businesses with product designers, software developers, QA engineers and project managers to support that work.
For World Usability Day, the challenge is practical: define what your product should help people accomplish, then examine how well the experience supports them—from the first interaction to the final decision.
