World Usability Day 2026: The Undo Step Belongs in the AI Demo
By Jonathan Anthony Burkett
Reporting and original analysis | November 11, 2026
A polished demonstration can make an AI feature look easy by showing a request followed by a pleasing result. The presenter asks for an organized set of notes, the interface rearranges them and everyone moves on. A more revealing demonstration would continue for another minute. What happens when the arrangement is wrong? Can the person identify what changed, keep a useful part and restore the rest? For World Usability Day 2026, the recovery step deserves a place on the stage beside the successful result.
The official World Usability Day website lists November 12, 2026, and the theme Designing a Sustainable Future with Ethical AI. The organization's 2026 event listing also connects that date with ethical AI. The following discussion is original design analysis prompted by that theme. It does not report findings from a product test or claim that a particular system meets an ethical, accessibility or environmental standard.
Use a task small enough to inspect
Consider a hypothetical feature that groups a person's project notes into folders. This is a deliberately ordinary example, not a description of an existing service. A useful demonstration would begin with a small, readable collection of notes whose original arrangement is visible. The person should be able to understand the starting point before asking a tool to change it. If the audience cannot inspect the input, an impressive output tells them little about whether the feature preserved what mattered.
The next step is a proposal rather than an immediate change. Show the suggested folder names and which notes would move into each one. Mark anything that the system could not place confidently, without requiring the audience to interpret a mysterious score. Let a person inspect the notes themselves. In this example, the practical question is whether someone can recognize a mistaken grouping before accepting it. A convincing explanation of the system's reasoning is less useful if the proposed action remains difficult to see.
Now introduce a deliberate mismatch. A note about buying materials might belong to one project even though its vocabulary resembles another. The person notices the problem and moves it back. This is a demonstration scenario, not evidence about how often a real system would make such an error. Its value is that it forces the presentation to show the correction process. A feature that looks helpful when everything goes well should also make a foreseeable mistake manageable.
Ask what undo actually restores
An undo label can conceal several different outcomes. Does it return only the latest note to its previous folder? Does it restore the entire arrangement? Does it also recover names that the system rewrote? A demo should show the scope directly, using the same small set of notes introduced at the beginning. The audience should be able to compare before and after. A reassuring button name is not a substitute for a visible account of what the button does.
Partial acceptance makes that question more interesting. Imagine keeping two sensible groups while rejecting a third. A single all-or-nothing reversal may force someone to choose between losing useful work and retaining a mistake. An alternative could allow the person to review changes individually, provided that the interface remains understandable. There is no universal design answer hidden in this example. The point is to make the tradeoff observable enough that a team can discuss it in relation to the task.
Recovery should also have an ending. After a person corrects a suggestion, show where the corrected state appears and how the person knows it has been retained. If the interface still displays an older summary, a user may reasonably wonder which version is active. A demonstration can expose that uncertainty without sophisticated measurement. Ask the person to explain the current arrangement in their own words. If they cannot tell what is now true, another animation or celebratory message is unlikely to resolve the underlying problem.
Make refusal an ordinary path
A useful AI feature should be demonstrable without requiring every suggestion to be accepted. In this hypothetical notes tool, the person might decide that the original arrangement works better. Show how to decline the proposal and return to ordinary work. Do not make refusal look like an error or a failure to understand the feature. A person who can make an informed choice has completed an important interaction, even if the system's proposed organization never becomes the final arrangement.
The same principle applies to language around the controls. Labels such as accept, edit and restore should refer to specific actions that the person can inspect. An interface should not require someone to decode a slogan at the moment they are trying to repair a mistake. Test the wording with the actual task in view. A question about whether a label sounds friendly is different from a question about whether a person can predict its result. Both can matter, but they need separate evidence.
For a design discussion, record three observations rather than a single overall impression. Could the participant identify the proposed changes? Could they correct the deliberately misplaced note? Could they explain the final state? These are suggested questions for a future evaluation, not results from an evaluation conducted for this article. They create a concrete agenda for discussion and help prevent a smooth presentation from standing in for evidence about everyday use.
The sustainability language in this year's theme also calls for precision. An easier recovery process does not, by itself, establish a reduction in energy use or environmental impact. Such a claim would need its own evidence and a defined comparison. Design teams can discuss wasteful repetition as an experience problem without attaching unsupported resource savings to it. Keeping those claims separate makes the ethical conversation more useful, because each assertion remains open to examination on its own terms.
A good demonstration leaves the audience with an understanding of the person's choices. What was offered, what changed, what could be refused and what could be restored should all be clear. World Usability Day provides an occasion to ask for that fuller sequence. Let the successful result have its moment, then show the mistaken result and the path back. The recovery step is where a promise of assistance becomes a practical relationship between a person and a tool.
Image note: Illustrative stock photograph of planning; not a World Usability Day event or an AI product demonstration. Credit: Barbara Olsen / Pexels. Source photograph; Pexels license.
#upstormed #jonathananthonyburkett #jonathanburkett #WorldUsabilityDay #EthicalAI
