AI is like nothing we’ve ever done before. It demands a new level of rapid, flexible, user-centered thinking and rapid adjustment, neatly expressed in this new process diagram.

In the beginning…

We’ve started our UX process diagraming with these very linear process diagrams, which were made specifically to make the management and clients more comfortable with the UX design process. Thus, early process diagrams like this one are linear, showing the neat value progression from definition to prototype to release:

Source: https://powerslides.com/powerpoint-marketing/web-design-templates/design-process-diagram/

Neat idea. 

Too bad that’s more or less complete bullsh*t. 

If we are honest with ourselves, this is what the process is like, “neatly” (ho ho!) captured by this diagram allegedly produced by a member of the original Apple design team (sorry, I can’t seem to find the reference for this anymore):

Of course, unless you’re a designer, that “creative mess” in the middle can be seriously intimidating. So we tend to make simpler versions of “the mess” such as this diagram describing the cyclical nature of one of my favorite UX design techniques: RITE (Rapid Iterative Testing and Evaluation):

Source: https://www.slideshare.net/usableinterface/rapid-testing-ct-uxpa-slideshare

However, more and more, these types of diagrams do not seem to capture the reality of designing AI-driven products and features. 

The main problem is that all these diagrams reflect the spoiled and lazy nature of the design industry. For many years, many of us had the luxury of not caring how our designs would be implemented. UX people neatly absolved themselves of the harsh realities of running a business. 

And to be honest, it was often true – technologies like Cloud Computing have conditioned us to care little about how something really runs on the back end, as long as there was a way to get from A to B and the implementation cost remained somewhat reasonable.

We had the luxury of focusing on the UX and leaving the implementation to the developers.

With the arrival of AI, all this is flipped on its head.

A New Way of User-Centered Thinking

To help explain how it is different and outline the new way of User-Centered Thinking, we’ve developed our AI-Inclusive UX Process, shown below:

The process is kicked off with an idea generated through a combination of Strategic Objectives and Market Analysis. The idea is then prototyped in a quick cross-functional brainstorming session conducted as part of the project kick-off.

Once the idea is documented and funded, it enters the core of this new normal: a rapid iterative problem definition and solution development cycle. This cycle is similar to that of traditional RITE but with one important difference: it adds a periodic AI-related “spike.” 

What the Heck is a Spike?

“Spike” in agile parlance is a quick, rough, proof-of-concept project designed to produce a piece of working code demonstrating that some idea is at least possible. The important part about the spike aspect is that it is super-lightweight and rough, almost the opposite of production code. In the case of AI, such a spike is best accomplished using a simple Python notebook setup for a very bare-bones I/O “experience.” It is the most basic proof of consent of AI aimed at answering the simple question: does this model produce the desired outcome? The purpose is to quickly nail down the problem definition and “proof of concept” a solution.

What is the role of Data?

The other part of the cycle is Data. Data is used to train and validate the AI Model. “AI” is actually made up of two interconnected parts: AI Model and Data. The two are closely intertwined and, for the purposes of UX design, can be thought of as a single system. 

The core of the process works like this: as we iterate the UX design starting the initial idea, we find new requirements for the AI model. When the AI model is iterated via a Spike, it requires new Data. Data that might have important parts missing or have detrimental biases or legal ownership constraints, which in turn affects the viability of the AI model, which in turn affects the availability of some aspect of UX desired by customers, requiring a design workaround and a cycle of new/modified ideas, which leads to re-spiking of the AI model, and so on.

Where is the Customer in all This? 

In the center. Where they always belonged. The whole process is centered around rapid iterative testing with the customer. If anything, using this process, customer feedback is brought much more forward in the development process. 

Why is this Change Necessary?

While the customer is still at the center of the process, with the advent of AI, it has become necessary to include technology in design solution iterations. That’s because, for the first time in many years, the tech is not 100% figured out. 

AI output is also really hard to mock up, so if we can save time by using a rough paper mockup PLUS a Python notebook to accurately tell the entire product story to the customer, that’s what we should do. Why? Because it is the most efficient way we can get accurate detailed feedback to tell us what to do next. And efficiency matters. A lot.

How does this affect the role of UX?

The role of UX has always been three-fold: remove the barriers for customers (usability testing of existing products), innovate and perform competitive analysis, and finally, act as a glue tying together customers, business, and technology. 

I suspect much of those responsibilities will remain. 

However, the role of the glue tying together customers, business, and technology will become increasingly more important with the inclusion of AI. That’s because this new technology is unknown and often unpredictable. This new AI-Inclusive UX Process is a testament to the need to research the technology itself, which has now become such an integral part of the experience. 

The other aspect that will gain prominence in the new “AI Normal” is UX-driven efficiencies and cost savings, driven by rapid prototyping, customer feedback, and accelerated lean decision-making.

The coming changes in UX roles are summarized below in this neat diagram:

Does this mean I have to learn about AI so that I can ask my data science teammates good questions?

Yes.

Final Hand-off to Dev

The final step of the diagram is the hands-off to dev. This is an important demarcation because it is new. In the past, production development often used to happen concurrently with design efforts. The new AI-Inclusive process diagram separates the exploratory “spike” development from the production efforts to secure and scale the system for production-level performance. From the UX perspective, not much changes there, other than making this hand-off an explicit step, separate from the customer-centered iterative nature of the “core” of the process.  

The other change is the feedback from the production department back into the core process, both in terms of customer feedback, model feedback, and newly generated data. It is a reminder right there in the diagram that the AI-driven product is never actually “done,” and the pace of change is only accelerating. 

Many More Changes to Come

AI is like nothing we’ve seen before. We are just beginning to explore the possibilities and requirements we face in the new normal. Continuous rapid adjustment will be the key to successfully leading AI-driven projects with UX.  

What did you think of the new process diagram? Let us know!

Greg & Daria

P.S. If you missed Greg’s plenary at the World IA Day, there are still FOUR upcoming unique learning opportunities at a discount:

Workshop: UX for AI: A Framework for Product Design (FINAL FEW SEATS — GOING FAST!)
03/16/24 UX Copenhagen, Copenhagen, Denmark

Helle has only a few tickets left for this very special conference. I’ve been trying to get to speak there for YEARS, so I’m very excited to teach my new UX for AI workshop at UX Copenhagen. Got friends in the EU? Please let them know! 

Panel: Writing in the Age of AI (25% off your ticket price CODE: WORDS2024AK25)
04/03/24 UX Salon WORDS 2024 (online)

Avi was kind enough to share this discount code with me: CODE: WORDS2024AK25 to get 25% off!

Workshop: AI in Search UX: A Framework for Product Design (on sale now)
04/22/24 Haystack Search Conference, Charlottesville, Virginia

This is perfect for East Coast peeps who are into Search UX. This will be a whole-day workshop focused on Search UX in the new world of AI-Driven products.

Workshop: UX for AI: A Framework for Product Design (on sale now)
05/21/24 UXLx, Lisbon, Portugal

UXLx needs no introduction. This conference is in May but is already almost sold out.

Please share with friends and colleagues. If you use the link below, I can send you my most popular book for free!