.png)
.png)
Social Media
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Copyright © 2023 Purpose UX, LLC.
Powered by Design with Purpose.
Powered by Design with Purpose.
More and more companies have added AI-powered features to their products, sometimes to great effect and other times to a resounding “meh.” In particular, designing good conversational AI experiences can be quite challenging, and many products miss the mark.
So if you’re building an AI feature or AI-based product, whether it’s enterprise, startup, or side-hustle, here are a few things to keep in mind.
General AI models are particularly bad at this, because they are attempting to serve a wide audience with incredibly diverse needs. Focus on who your users are and what they are trying to accomplish with your product so you can give the right assistance.
Avoid verbose text instructions — quick prompts, one-click actions, and other shortcuts can help people understand how to get what they need.
Some users are happy to go back-and-forth with a conversational UI, but most are trying to accomplish a task as quickly as possible. Bundling next steps, follow-up actions, etc. with a typical response can relieve some of the burden of knowing what to do next.
Don’t worry about this feeling “artificial” — people know they are interacting with an AI and are generally happy to have explicit options.
Old-school text-based adventure games were notorious for how opaque the gameplay could be as you would have to guess what commands to type in to get the desired result.
Picking color palettes, making precise selections or edits, and other similar actions can be done by AI Agents, but that doesn’t necessarily mean they should *have to* be done by AI Agents. Where appropriate, provide explicit controls, either within the conversational UI or as a fall-back.
One of the big advantages of LLM interfaces is that they can handle long lists, paragraphs of instruction, and other detailed, long-form input. For many people, typing out all that text is a bad experience (especially on mobile!)
Make sure to support text-to-talk dictation and or other ways to ease the burden of input.
It’s become a meme with how many products are incorporating AI features, and in some cases, aggressively pushing people to use them.
Don’t go along with the crowd just because a stakeholder says so or because of a vague feeling of being “left behind.” Understand your audience, their needs, and what they use your product to accomplish. If AI can support those needs in an obvious way, great, go for it! If the signals are mixed, consider a low-overhead pilot or research study to tease out where the value is. And if AI seems like a distraction or annoyance, be prepared to have some tough product conversations about where your priorities should be.
Also understand that many audiences (especially in creative industries) are strongly opposed to AI for ethical, aesthetic, or other grounds. Adding AI to your product without understanding the landscape can lead to customer backlash and damage to your brand. Again — do your research!
It's tempting to think that a conversational AI is automatically a good experience - this is not true! The same principles for designing traditional interfaces apply here as well - understand your user's needs, illuminate the path to completion, and provide good affordances for what actions can be taken. Do you know any other good principles for designing conversational AIs? Let us know!
