By Gabriel Luján.
Since the first half of last year, I have been working on products that make extensive use of artificial intelligence, an experience that has been extremely enriching as a product manager - and designer by vocation.
During this time, I have used more AI products than I can count and followed the evolution of the discipline in all its aspects: Engineering, design, data and applications.
Here, I compile what has become established as experience conventions. But what exactly is that?
Conventions, in UX, are patterns that facilitate the distinction of visual elements and, consequently, reduce the cognitive load needed to use whatever is being built.
For example, a "STOP" sign is red with white capital letters because it was established that way. Finding a blue "stop" sign with black lowercase letters would take more time to interpret than its intended use allows. You would probably ignore it and your car would pay the price.
In UX, this happens all the time. How many buttons that didn't look like buttons did you take time to notice? How many poorly designed text input fields did you struggle to find?
This is an article about usability, which doesn't mean it's an article whose content is only interesting to designers. If you are - or intend to - build a product that uses artificial intelligence, some of the conventions here might be useful to you.
Whether as a provocation to those responsible for your application's experience - or as a proactive gesture to improve the product with your own hands.
Usability is everyone's responsibility.
Although I said this is an article about conventions, I mix in some concepts that could be considered heuristics. However, my focus here is on applying these "rules" in creating recognizable elements, expected by users and with perceived results.
When we build a product, we tend to base conversations on objectives before listing solution hypotheses. Although this is an excellent exercise, we need to acknowledge that everyone enters a co-creation process with some ideas of what is possible.
When someone says the goal is to increase the number of users, no one considers the hypothesis of launching a satellite so that isolated villages without internet access can start using the product. We are limited by the product development paradigms we know, and that's okay. They keep us grounded and the conversation productive.
Since the advent of mobile applications, no disruption this significant had occurred in software development paradigms. AIs became a hammer looking for nails to hit. Established products began to imagine how they could improve by using LLMs. Others thought about how to create a product where AI is the core of the business.
With new capabilities came new experience needs and, consequently, exercises in how to reduce the friction in using these applications.
There are a lot of "flavors” of AI products. AI First, AI Enhanced, Invisible AI applications.
Here, we'll build together an AI First product – a simple and powerful chat, but the same applies to any other type of product.