Designing for AI: patterns for conversational interfaces
Chat is the default interface for AI products, but a blank text box is rarely the best one. Here are patterns that make AI features easier to use.
Creative Pouch · · 1 min read

When a product adds AI, the first instinct is usually a chat window. It’s flexible and familiar, and it’s often the wrong choice. A blank text box asks users to know what’s possible and how to ask for it.
Good AI interfaces guide people. Here are patterns that work.
Start with suggestions, not a blank box
Show a few concrete starting points: “Summarise this page”, “Make it shorter”, “Suggest three headlines”. Suggestions teach users what the feature can do without a manual.
Put AI where the work is
The best AI features often don’t look like chat at all. A “rewrite” button next to a paragraph, an “auto-tag” action in an asset library, a smart default in a form field. Bring the model to the task instead of making users go to the model.
Show the model is working
AI responses take time. Stream the answer as it’s written, show progress for long tasks and let users cancel. Silence for ten seconds feels broken. Visible progress feels fast.
Make output easy to edit
Treat every AI result as a draft. Let users edit inline, regenerate one part, or undo in one click. The less it costs to correct the model, the more people will trust it.
Be honest about uncertainty
Label AI-generated content clearly. Where accuracy matters, show sources or let users check the inputs. Don’t present a guess with the same visual confidence as a fact.
Remember the conversation
If users have to repeat context every time, the feature feels dumb. Keep relevant history, show what the model “knows” and let people clear it.
Design the failure states
Models misunderstand, refuse and get things wrong. Design those moments as carefully as the happy path, with clear messages and an obvious next step.
The goal isn’t to make AI feel human. It’s to make it feel useful.


