As Artificial Intelligence moves from an idea for a conversation to a hidden part of the software we use every day a new problem has come up: the token budget. In language models a token is the basic unit of processing text. Every time you click a button search for something. Get a suggestion to complete what you are typing it costs money and computer power. As the people who design products our challenge is not just about making things look good anymore. It is about helping people focus on the things given the limits and costs of Artificial Intelligence computing.

Big tech companies, like Notion and Figma and Vercel are figuring out this area. They are finding that the best experiences for users are made by designing within the limits of what's possible and turning those limits into good user interface design.

"In the age of Artificial Intelligence good design is directly related to using tokens. The elegant interfaces are the ones that give users the most help while using the least computer power."

The Good and Bad of Artificial Intelligence Integration

Adding Artificial Intelligence that can generate things to applications has big advantages and disadvantages. Understanding these is necessary for designing systems that're helpful to users and good for businesses:

// THE GOOD THINGS

  • Artificial Intelligence can make interfaces that change to fit what each user needs and how they work, in time.
  • It can make it easier for users to interact with the system by turning actions into simple voice commands.
  • It can help users start the process by giving them templates and drafts right away.

// THE BAD THINGS

  • It can be overwhelming for users with many options and suggestions which can cause fatigue and visual clutter.
  • It can cost a lot of tokens, which can make the cost of using the application high for the business.
  • It can make the system slow by making users wait for responses, which can break their concentration and focus.

Designing the Interface that's Aware of Tokens

To build Artificial Intelligence systems that are responsible designers are using several strategies that take tokens into account. First we design systems that need users to intentionally trigger the Artificial Intelligence. Of running it automatically we require users to do something like press the space bar in an empty block in Notion. This makes sure that tokens are only used when the user really wants to use them.

Second we focus on showing users the Artificial Intelligence capabilities a little at a time. Of showing a big chat box everywhere we put small agents directly into the existing interface elements like a spell check or a smart search bar that knows the context. This prevents overwhelming the user. Also reduces the number of tokens used by narrowing down the context of the prompt.

The Future of Design that Puts Resources First

Designing for Artificial Intelligence requires an understanding of how software is made. By seeing token limits, as a challenge rather than a technical problem we can design products that feel fast, clean and intuitive which shows that good user experiences are always built on smart system engineering.