What new challenges does interface design face when AI becomes part of a product?

From predictable actions to shared judgment

In traditional interfaces, users generally expect a clear result when they click a button. AI features may need to interpret input, take time to process it and generate varying results. Interfaces therefore need to help users understand what the system can do, what it is doing and whether the results need to be checked.

Building on public discussions at CES about trust and transparency in AI content, this article offers independent interface design recommendations. A product's specific capabilities and data handling should be verified against its actual implementation.

Explain capabilities and scope first

Describe the feature through specific tasks at its entry point, such as "Organize questions based on the materials provided," and state the applicable materials, versions or other conditions. Avoid unverifiable claims such as "understands every need," and do not let conceptual effects shown in demonstrations be mistaken for capabilities already delivered.

When asking customers to provide information, explain its purpose and why it is necessary, and handle it in accordance with the company's actual privacy notice and applicable requirements. If customers do not wish to provide additional information, they should be able to find out whether there are other ways to complete the basic task.

For questions involving quotations, contracts, medical matters or other issues requiring professional judgment, the interface should provide appropriate ways to seek further verification. It should not present uncertain generated content as a firm commitment.

Make status and results understandable

During processing, indicate whether the task has been received, whether it is still in progress and whether the user can stop it. Display progress or save status only when the system actually knows it; when it cannot be confirmed, say so honestly.

Results should clearly distinguish content provided by the user, suggestions generated by the system and facts that have been confirmed. Where appropriate, provide sources, applicable conditions or ways to verify the results to help users decide whether they are suitable for use.

When information is insufficient, an answer is uncertain or an operation fails, explain what is known and what to do next, rather than masking the problem with vague error codes or false success messages.

Preserve user control before key actions

When sending content, modifying data or carrying out actions, let users review, edit and confirm at appropriate points. After a user cancels or changes a selection, the interface should reflect the system's actual behavior. Do not promise an undo function that does not exist.

If the system needs to pass information to a human adviser, share only the necessary content within the scope of appropriate consent and permissions, and allow users to review the summary. Mark information that customers have not explicitly provided as awaiting confirmation, and do not treat system inferences as facts.

Maintain a clear route to human support or another effective option, and display actual service arrangements. If a transfer fails, explain whether the request was delivered and how to try again.

Validate with tasks and exception scenarios

Invite target users to complete real tasks and observe whether they understand the system's capabilities, can identify errors, edit results and find help. Cover practical conditions such as mobile use, keyboard operation, network interruptions and outdated materials.

Record the conditions that trigger each issue, its impact and who is responsible for addressing it, then retest after changes. Limited testing can help identify barriers but does not represent all users. Qingmei Future can assist with product interfaces and user experience research; the company is responsible for confirming technical implementation, information accuracy and service commitments.

During testing, check states such as processing in progress, results requiring verification, an action already executed and a failed request separately. Ask participants to describe what they think happened, then compare their understanding with system records to identify where messages and actual behavior differ.

For high-impact actions, confirm whether customers can check the target and scope before execution. If some actions have already taken place despite cancellation, show the status accurately and provide remedies that the company actually supports.

After launch, regularly check that explanations, features and support channels remain accurate. When features change or materials are updated, revalidate the affected tasks rather than only revising the interface text.

Discussion Background and Scope

CES · AI or Human? Trust, Transparency, and Co-Creation in the Age of Intelligent Content · 2026-01-07. Official page date: 2026-01-07. The source discusses authenticity, transparency and accountability in intelligent content. It is background reading, not evidence for each practical recommendation in this article. The recommendations are Qingmei Future’s independent analysis.

Related Services and Content

Explore digital interface design　Explore industrial design services　Discuss product experience design