Do AI chatbot plugins spell the end of UX as we know it?

In a recent blog post, our CEO Martin Buhr discussed the possibilities afforded by ChatGPT plugins in the world of API management, especially around natural language interfaces. This is a very exciting area, as it opens up AI assistant-style possibilities for all aspects of the systems supported with API management platforms like Tyk. What if you could take all the complex settings fields for an API and replace them with a simple conversation about what you want to achieve? It’s intoxicating stuff.

Selfishly, though, my thoughts turned to UX, and the implications for a team researching and designing user interfaces. What do these advances mean for us?

Why might ChatGPT plugins be a bad thing for UX?

What if we replaced every interaction in the system with a simple chat bot window? Describe what you want done, and the bot does it. Ask for a custom chart, and the bot draws it.

Interaction design would become irrelevant, as there would be only one form of interaction:

      –  “Tyk, show me the policy details for this API.”

      –  “Tyk, create a new graph that combines all the APIs on this environment.”

      –  “Tyk, make sure the API documentation for this product is up to date.”

And interaction design – on both the research and design sides of UX – is a huge part of what we do.

Researchers use product analytics (such as heat maps) to understand how users are interacting with systems, and how successful they are at achieving their Jobs to be Done. Designers spend hours carefully crafting interactions through a combination of user psychology, industry standards and sheer creativity.

Strip that away and replace it with a chatbot interface, and there’s suddenly a lot less for us to do.

On the other hand, might it be a good thing?

In all my doomsaying, I’m ignoring the very thing that UX professionals should be good at – understanding human nature.

When examining metrics on the use of a system, it’s one thing to know what the numbers tell you – “users aren’t clicking this bit here” – but it’s quite another to understand why. Good research is predicated on understanding and accounting for human nature, digging down beyond a surface interpretation of the facts to get to the truth.

An AI assistant could help us with design, too. For instance, we have a fairly well-defined UI style – we could just tell our AI what users want from a page, and the algorithm could use what it has learned about our existing software to create a simple prototype of the page design. It’s not inventive or creative, but that’s not what AI is really about, at least not yet.

That ‘starter for ten’ UI could be just the object we need to discuss the basics of an interface, preventing designers having to do repetitive and unfulfilling tasks and leaving creativity to flourish where AI cannot, which is in taking the simple UI and adding delightful interactions to elevate it above the ordinary.

It’s about more than just design and execution

Tyk’s Head of Engineering, Leonid Bugaev, shared some excellent wisdom (as we have come to expect from him!) when I was discussing these thoughts:

The overall quality of UX work done via ChatGPT, if you are a very small team (or single person), should raise the baseline, help improve productivity, but not replace creativity just yet. There is a genuine use case for junior/mid-level UX/UI designers, and freelancers, as it will be able to “execute” to a good level.

There are already apps which have basic collections of UI elements, and you can enter prompts like: 

      –  Build me an analytics dashboard, with a line chart, and the following filters, and a searchable table below it”

And it will build you a good-looking UI following good UX standards. But designing good products is not just about having good UX, but understanding your users, and solving their exact pains in an efficient way. It’s not about execution itself. So if you are creative at generating prompts, you can get creative results out of AI tools.

In the past, a researcher’s core skills would include ‘the ability to search effectively in Google’ (not as straightforward as it seems). So the next big skill for all professions will now be ‘generating effective prompts for ChatGPT’. 

A wise man, indeed. 

Good professionals might still use the same tools as others, but it’s how they extract the absolute best from those tools that will set them apart as the technology becomes ever more prevalent.

So, is AI a threat to UX teams or not?

Overall, it’s hard to know right now. ChatGPT is a very good machine learning engine tied to a very good inbound and outbound NLP engine, but that’s all it is. It doesn’t feel like the leap to AI that always seems to be just a few years away, rather like commercial scale nuclear fusion reactors. That makes it less impressive and less scary from a layman’s perspective. It’s not going to steal all our jobs… yet. In fact, it might just elevate our roles to new heights.

But, and it’s a BIG but, we need to stay ahead of the curve, rather than waiting to see what happens. We need to be exploring the opportunities provided by machine learning and AI as soon as we can, and be ready to pivot our roles to take advantage of its ability to do the mundane, while we do what only humans can do: invent the incredible.