AWS Builder Center

AI-Controlled Interfaces: One Interface, Different Experience

An exploration of how AI can adapt a single underlying interface to different users, tasks, experience levels, and business requirements while preserving user freedom.

AI-controlled interfaces can take a very different approach from traditional application design. Instead of creating separate screens and workflows for every type of user and task, one interface can potentially support the entire application. AI determines how that interface should look and behave for the particular user and the task at hand. The same underlying interface can therefore look very different to different users without becoming a different application.
When the interface starts, the AI uses a defined UI Model as its foundation and combines it with a User Model. The User Model can group users according to their role, experience, usage patterns, and other relevant characteristics. As the user becomes more familiar with the application, AI can recognize the user's learning curve and gradually introduce more advanced functionality when it becomes relevant. At the same time, the interface can adapt its presentation—for example, changing the position of fields or deciding which menu items should be most visible for that particular user and task. The goal is not to restrict the user, but to allow the interface to become more capable as the user's experience grows.
At the business level, a Business Model can define how the interface should support different user groups, how users may move between groups, and how the application can evolve as the business grows. It can also define legitimate prevention rules for situations the business has identified as problematic. These rules should be created and controlled by the business, not by the AI. AI can recognize when an action matches an existing rule, but it should not invent new restrictions simply because it considers an action undesirable.
This separation becomes especially important when the interface adapts continuously. AI can personalize the presentation, recognize highly likely user actions, and make frequently used functionality easier to reach. At the same time, hiding a menu item should never mean removing access. A user should always be able to request and see all functionality they are legitimately authorized to use. The AI controls how functionality is presented, while the underlying models and business rules control what functionality exists and what the user is authorized to do.
This is not yet a fully implemented AI-UI model. It is an architectural direction that is worth exploring. In the second part, we will look at some of the key functionality and logic required to make this type of interface practical, including how the AI, UI Model, User Model, and Business Model can work together.
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article