What should be different about the UX of AI products? - (1) Automatic trap
TeamLilys
23
2
I would like to share one by one what I learned and felt while creating three AI products: Vrew, Rutton, and LilysAI :)
With the incredible advancements in artificial intelligence, there is now an opportunity to ‘automate’ many human intellectual tasks.
But ironically, when our team discusses the UX of a product, we often talk about not falling into the “automatic trap.”
When you ask AI to do something, it's rare for it to get a perfect score. Most of the time, it ends up getting around a 70.
You might think that it would be meaningful if you could just create a draft with 70 points, since it would save a lot of time. However, as you follow the user's workflow, you will often find yourself having to destroy the 70-point result created by AI and create 100 points from scratch to achieve your desired goal.
For products like this, users initially react by saying, "Wow, this is amazing," but in reality, they don't reach a usable level, so retention isn't very good.
We call this phenomenon the 'automatic trap'.
So how should I fill in the remaining 30 points?
A lot of things could be solved with the engineering and UX layers.
Let me give you an example using our products.
LilysAI is a video summary service.
There was an issue where the summary was sometimes inaccurate or overly abbreviated, making the reliability questionable. When I used the existing video summary service, I felt inconvenienced because there was no action that the user could take when the reliability was questionable, so I ended up having to watch the video again.
So we designed the user experience so that 1) the original script that serves as the basis for the summary is placed right below the paragraph, and 2) the video player is always fixed to the left so that the video corresponding to the summary and the summary notes can be viewed together.
As we continue to improve our services with this product philosophy, D+1 Retention has also continued to increase from the 30% range to the current 56%.
One of the most common comments from LilysAI users is, “LilysAI may not seem like a flashy service, but it seems to be the AI product they use most often.” I really like this comment.
Going forward, we will focus more on building a usability layer that truly solves users' problems rather than just fancy-sounding technical keywords.
If you are a product maker who is interested in our AI product philosophy and would like to learn more, please apply for the position we are hiring for. We are always looking to meet like-minded individuals.