2023 was the year of many things, but more dramatically, the rise (and sometimes fall) of AI products. As a designer in tech, in a world where every new startup tries to automate some part of our day-to-day work, you might ask yourself, “Is designing AI products any different than any other digital product?”
Wrapping up a year of designing AI-driven creative tools, my short answer will be “Not really,” but there are some key learnings I picked up along the way.
Context and copy DO make the difference.
When working on a technology-driven product, many users are not yet familiar with the technical terms and codes the industry uses. So, to be clear on which input is required from your users, using terms like “prompt,” “algorithm,” or “output” might not be the way you want to communicate with your first-time users.
Instead of making them guess the meaning, it is way better to describe what action is required to take and why. For example, instead of the word “prompt,” using a simple “description” fosters more familiarity with the action and less confusion. Use context or examples as a way to ask for more details, direction, and clarification.
Give users a place to start instead of a blank page.
As if facing a white sheet of paper or a document isn’t hard enough, facing a completely new tool with zero experience is way harder to start from. Whatever business solution or support your tool provides, users don’t always come to you with clear ideas in their minds.
Sometimes, they try things out, and ideas follow after. As one of my brilliant leaders once said, “People don’t just show up to the restaurant the first time and expected to order right away. You have to suggest your options first before they make any decision.” I think this describes the concept of onboarding experience to an AI tool perfectly.
Bad conversion isn’t always a design flaw. Be kinder to yourself.
Sometimes, especially at the beginning of any “AI in training” stage, the quality of the result just isn’t good yet, especially if it’s custom-made and not an API integration of an already fruitful algorithm.
Before your baby AI learns enough, your users might not yet be satisfied with the AI’s output. It does not always mean you have a “bad UX.” It can be, but not always. Instead of redesigning the same flow over and over again, the issue should be identified by data and research before jumping into a new redesign scope. This is where the most collaborative approach with research, data, management, and design does the magic.
Thank you for reaching out here! 🎉
At the end of the day, AI or not, we design for humans and human experiences, so empathy is all you need to design better products.
Have an exciting journey to share? Leave a comment!
& don’t forget to stay in touch on Linkedin!