Beyond the Chat Box: Choosing the Right Interface

When a team decides to "add AI," the first design that appears on the whiteboard is almost always a chat panel in the corner of the screen. It's familiar, it demos well, and it can technically do anything. It's also often the reason the feature goes unused: a chat box asks users to stop what they're doing, figure out what to ask, type it out, read a paragraph back, and carry the result to wherever it was needed.

Most of the best AI features don't look like chat at all. They're a pre-filled dropdown, a button next to a selection, a summary already waiting at the top of a page, or a search box that simply returns better results. In this lesson you'll learn the main interface patterns for AI features, a short set of questions for choosing between them, and how to build two of them in Deskmate.

What a chat box asks of your users#

Chat feels effortless because typing a question is easy. The hidden costs sit on either side of the question:

• Knowing what to ask. A blank box gives no hint of what works. Users who don't already know the feature's strengths try something vague, get a vague answer, and conclude it isn't useful.
• Putting the context into words. The user is looking at a ticket; the chat panel doesn't know which one unless you wire that in, and even then the user has to describe what they want done with it.
• Reading prose to extract a result. If what the user needed was a category, a date, or a corrected sentence, a paragraph of explanation is overhead.
• Carrying the result back. Copying a drafted reply from a chat panel into the reply editor is a small chore that happens hundreds of times a day.

None of this makes chat bad. It makes chat expensive per use, which is fine for open-ended, occasional tasks and wasteful for narrow, frequent ones. The question for every AI feature is where it should live so that the cost of using it is lower than the value it adds.