I read an article last week by a wannabe influencer predicting that AI would bring about the UX apocalypse, and that designers everywhere should pack their things and look for employment elsewhere. This isn't the first article of its kind that I've come across recently, but what worried (and surprised) me about this one in particular was that it was published by a well-respected UX blog.
Click-bait nonsense like this is dangerous. If you've ever had a conversation with an investor, you'll know how quickly something someone read somewhere becomes gospel, and before you know it, it's an established factor in funding decisions everywhere. People lose their jobs over this stuff.
So what was it about this article in particular that triggered me?
Well, the claim was that Artificial Intelligence doesn't require any kind of UI, given that it's usually deployed in a chat-based wrapper (ChatGPT, Claude etc). And since AI is rapidly becoming the dominant computing paradigm, the need for user interfaces in general will quickly disappear over the next few years. This will eventually result in a profusion of UI-bereft chatbots, rendering the UX design discipline effectively redundant.
The implication was that UX design was on the road to extinction, and that anybody who wanted to safeguard their careers should start re-training immediately to become Prompt Engineers. Of course, all of this is absolute nonsense, and demonstrates the author's fundamental misunderstanding of both UX, and of Artificial Intelligence.
There are two major problems with this prediction: Firstly, it assumes that UX is purely about user interface design (it's not). Secondly, it assumes that chat-oriented design patterns are the only way to interact with AI (they're not). In this post, I'll explain why neither of these assumptions are even close to the truth, and get into some detail as to why I believe the AI revolution will actually re-invigorate a stagnant UX industry and create new opportunities for creativity and innovation in digital product design.
UX ≠ UI
When most laypeople think of user experience design, they tend only to consider the user interface element of it. They think about the usability aspects of it, and maybe even the graphic design element of it, but their perception of UX usually stops at pretty boxes and arrows. In reality, it goes far, far deeper than that.
Proper user experience design encompasses every aspect of the journey - not just the pixels a user moves around on the screen to get a job done, but their entire experience from beginning to end - including the bits that happen in the background that they don't manipulate directly. In the new AI paradigm, that means designing how the machine should interpret and interact with a user's input, and how it should convey its response back to them.
Which brings me neatly on to the next point…
AI ≠ Chatbot
Artificial Intelligence - and our interaction with it - is a new technology paradigm that isn't fully understood yet. It's still quite experimental in nature, which means that standard design patterns have yet to be established and homogenised. If anything, this creates an exciting new frontier in human-computer interaction that hasn't existed for a long time.
Most of the AI tools we see on the market these days fall into one of two categories:
- General purpose chat interfaces that are deliberately simple
- Experimental (and often gimmicky) enhancements to existing products
It's important to remember that these are not the finished article. They're early-to-market prototypes, still attempting to prove their quality and value to a mass market without being tied to any specific use-case. Chat interfaces are by their nature ubiquitous and simplistic, which means they're very quick and easy for engineers to deploy and iterate without having to spend valuable time on the nuances of UX.
Ubiquity is the key. They absolutely have to work for everyone everywhere all of the time, because AI only works at scale. If they were too niche, they wouldn't have access to sufficient data to evolve and be useful. Dramatically simplified design patterns like chat and search were the obvious choice for early AI engineers looking for an easy HCI paradigm on which to hang their natural-language technology. The early movers did it first, and everyone else just copied a pattern that appeared to work.
A new computing paradigm
In 2023 Jakob Nielsen declared that Artificial Intelligence was the third computing paradigm. The first was what he called “batch processing”, where a long continuous string of sequential commands were typed or cut into punch cards and fed into a massive computer the size of a small building to process all in one shot - often overnight or over a period of days. The programmers had to think of everything they wanted up-front, and if the output wasn't exactly what they were looking for, the whole process would have to repeated again from scratch.
The second paradigm was a little more forgiving - what Nielsen called “command based interaction” - and came about in the 1960s. In this paradigm, commands and responses were passed back and forth between operator and computer over and over, until a satisfactory result was reached. This is what we've been accustomed to for over half a century. First, we did it using a command line (where a programmer would type in code, and the computer would print the response back on-screen), then later through a graphical user interface - starting with keyboard and mouse, before evolving into touchscreens and voice input.
None of these paradigm shifts spelled the end for those who created the software that powered them. Engineers didn't suddenly find themselves in the queue at the Job Centre when graphical user interfaces came out. If anything, the engineering role diversified and exploded as a result. As soon as ordinary folk were able to use computers, the engineers who created them didn't suddenly find themselves redundant - their focus simply switched from operating the machine itself, to creating tools that made it increasingly easier for everyone else to do so.
The simplistic chat and search style interfaces we see today are only the first crude prototypes of an exciting new paradigm in computing. We've got an awfully long way to go yet before we find the most fruitful and efficient ways of getting the best out of AI.
With the ubiquitous chatbot design pattern, we've effectively rolled back to that first “batch processing” paradigm. Yes, the underlying tech is smarter and faster than it was the first time around, but our interaction with it is still essentially the same - the user needs to think of everything up-front, and get it perfectly right in a single hit, or they'll have to go back and start over. And this is what makes it completely inaccessible to the majority of people, who struggle to articulate fully what they want or need.
This has led to the emergence of a preposterous new job title - the Prompt Engineer. This implies that, in order to unlock the power of the technology, we must rely on a skilled technician who knows the correct syntax. For those of us old enough to remember it, this all feels very familiar - the AI chatbot has effectively become the new command line.
So where do we go from here?
The challenge for us all in Product Design will be to integrate Artificial Intelligence seamlessly into established UX design patterns. And I mean *fully* integrated - not simply bolted on with a twinkling magic-wand icon. Truly integrated, to a point where it becomes an invisible part of the existing machine, delivering genuine value to everyone, without them having to know instinctively what to ask and what language to use.
It will not be the one-size-fits-all approach that we're seeing right now. Our interactions with AI will become as diverse as the use-cases they're deployed into; the best and most fruitful implementations being guided experiences, where the user is assured helpful and truthful responses through carefully placed guard-rails.
The most effective AI implementations generate detailed prompts behind the scenes based on a user's input, giving the model all the context it needs to simulate intelligence effectively - carefully engineered to provide value without relying on a user choosing the right syntax.
And this is where the new UX challenge will be found - where it has always been - in the masking of technological complexity to ease the creative collaboration between man and machine.
