AI is changing the value of UX

Much of the conversation about AI and UX has focused on productivity. Can AI generate wireframes? Summarize research? Write requirements? Produce prototypes? Help designers work faster?

Of course it can.

Greg makes a useful distinction between AI for UX—using artificial intelligence to accelerate UX work—and UX for AI, the work of designing AI-powered products and experiences themselves.

The distinction matters because the first category is becoming increasingly accessible. If one designer can use AI to produce an interface more quickly, so can another designer. Productivity improvements are valuable, but they aren't necessarily durable sources of professional differentiation.

Greg believes the more consequential opportunity lies in moving upstream.

UX professionals can help determine which customer problems are worth solving, translate customer needs into product capabilities and understand enough about AI technology to participate meaningfully in decisions about how those products should work.

That means the role begins to look less like interface production and more like AI product design.

Mike saw a parallel in UX research. Researchers are understandably interested in using AI to make their existing work more efficient, he said, but there has been comparatively little discussion about a more disruptive question: How does AI change the rules of a good experience?

An AI product may not primarily be a collection of screens. It can be a conversation, an assistant, an agent or an evolving relationship between a person and a system.

That requires different design muscles.

When building gets cheaper, discovery gets more valuable

There is an understandable temptation to conclude that faster development reduces the need for deliberation.

Why spend days researching an idea when AI can build a prototype in hours?

Greg reaches the opposite conclusion.

As the cost of creating software falls, the importance of deciding what to create rises. A team capable of producing five prototypes instead of one has not necessarily become five times smarter. It has merely acquired five opportunities to be wrong.

Greg advocates an iterative process he calls a “snowball sprint.” Teams begin with something small, validate it with customers and gradually roll additional functionality into the product as they gather evidence that the concept is heading in the right direction.

“It’s an iterative process with the customer in the middle, not AI,” he said.

That last distinction deserves attention.

AI product development can easily become technology-centered rather than customer-centered. A new model appears, teams discover an intriguing capability, and the search begins for somewhere to deploy it. The technology becomes the premise; the customer problem becomes an afterthought.

Greg's approach reverses the order.

Before substantial building begins, he recommends framing the use case, understanding the available data and examining value, ROI and risk. In the interview, he describes three exercises—storyboarding, a digital twin and a value matrix—that layer together to interrogate those questions.

Only then does rapid AI prototyping become truly powerful.

The point isn't to slow innovation. It's to focus on its aim.

Fast failure has a hidden cost

Mike raised another uncomfortable consequence of AI product development: the mythology of “failing fast.”

Agile software development helped popularize the idea that teams could experiment, learn and correct mistakes quickly. AI pushes that logic even further because creating and revising products can happen at remarkable speed.

But speed can disguise waste.

“Who likes doing rework?” Mike asked. For companies operating with limited money, time or organizational patience, there are only so many failed iterations available before experimentation starts looking like poor judgment.

With AI, the stakes can also extend beyond wasted development resources.

AI systems can produce incorrect outputs, misunderstand instructions or take actions with consequences that conventional software does not. That makes responsible AI design and AI risk management inseparable from the user experience.

Greg compares an LLM to something that must be carefully aimed: giving the system a goal does not guarantee it will pursue that goal in the way its designers imagined.

The old product question was often: Can users figure out how to use this?

AI product teams must ask considerably more.

What happens when the system is wrong? What happens when it is confidently wrong? What can it act upon? Can the action be reversed? When should a human intervene?

These aren't merely engineering questions. They are experience questions.

AI ROI needs to include the cost of being wrong

That leads to one of the episode's most useful discussions: AI ROI.

Companies naturally want to know whether an AI investment will increase productivity, reduce costs or generate revenue. But Greg argues that calculating the value of an AI system also requires examining the consequences of its decisions.

He describes using a value matrix built around the familiar categories of true positives, false positives, true negatives and false negatives. The important step is attaching real-world consequences to each outcome.

“If it got it right, great. It saved some time,” Greg said. But teams must also ask, “What happens if it got it wrong?”

The answer won't be symmetrical.

A wrong restaurant recommendation might cause mild annoyance. An incorrect action in cybersecurity can carry substantially greater consequences. The UX for AI challenge is therefore not simply maximizing accuracy. It is understanding which errors matter, how much they matter and what safeguards the experience needs because of them.

And Greg insists those answers cannot be found entirely inside a conference room.

“You cannot come up with these answers just by sitting and drinking lattes around the office,” he said. Teams have to talk to customers and understand what actually happens when the product succeeds or fails.

Customer discovery becomes risk discovery.

The UX profession may need to rediscover itself

For UX practitioners worried that AI will diminish their role, Greg offers an argument that is simultaneously reassuring and demanding.

The profession's real value was never Figma.

Design tools became so intertwined with design work that proficiency with the instrument could sometimes be mistaken for the purpose of the profession. AI is now breaking that association.

Greg uses a superhero analogy: Thor is not the god of hammers. The hammer focuses his power; it isn't the source of it.

For designers, he argues, tools work the same way.

The enduring capabilities are empathy, curiosity, problem-solving, facilitation and the ability to turn ambiguous human needs into workable solutions. What changes with AI-powered products is that practitioners also need enough technical fluency to understand concepts such as retrieval-augmented generation, memory and AI agents—and enough confidence to help shape the product rather than wait for requirements.

“You need to grow into your space as a leader,” Greg said.

That may be the more useful way to think about AI's effect on UX careers. Automation is not simply eliminating tasks. It is changing which layer of the work commands value.

Execution becomes cheaper. Judgment becomes more expensive.

Customer focus is still the competitive advantage

Near the end of the conversation, Mike asks Greg to look several years ahead and identify what will distinguish companies that succeed with AI.

After more than an hour discussing rapidly changing technology, Greg's answer is strikingly traditional.

Customer focus.

“I think it’s what always separated them,” he said. “It’s the strength of the brand built on relationship you build with your customers. It’s this long-term loyalty and long-term customer value.”

Perhaps that is the central irony of UX for AI.

The technology is changing extraordinarily quickly, yet the principles most likely to keep companies oriented are stubbornly human ones. Talk to customers. Understand the problem before proposing the solution. Test assumptions. Consider consequences. Measure value in the real world. Use technology to increase human capability rather than merely demonstrate technological capability.

AI gives product teams an extraordinary new engine. UX, research and product strategy still have to decide where the vehicle should go.

For Greg, that creates an obligation as much as an opportunity for the people who understand customers to become more involved in shaping AI and not retreat from it.

“The field needs you to stand up and be counted,” he said. “If you’re a product leader, if you are a UX leader, if you’re a researcher, you need to take an active part in studying this.”

Related resources

• Leveraging AI in UX design. This on-demand webinar explores AI as a strategic tool in the design process, including principles for designing AI capabilities and integrating AI into UX methods. It complements Greg’s distinction between simply using “AI for UX” and developing the skills required to design AI-powered products.
• How to test AI: A practical guide for evaluating AI user experience and product design. This blog post focuses on embedding AI UX research into product development to reduce risk and validate AI-enabled experiences, including agents and copilots. That closely reflects Greg’s argument for putting customers at the center of rapid AI prototyping.
• UX research for AI: building trust in experiences. This Insights Unlocked episode with Microsoft senior UX researcher Priyanka Kuvalekar explores why evaluating AI requires going beyond traditional usability to consider trust and emotion. That connects directly with Mike and Greg’s discussion about how AI changes the rules of a good experience and why traditional usability testing alone is no longer sufficient.
• How to test AI features: rethinking AI usability testing for conversational experiences. This blog post argues that AI features shouldn’t be evaluated like static interfaces because they behave more like conversations and relationships. It’s an especially natural follow-on to Mike’s observation in the interview that UX professionals increasingly need to think about designing and testing conversations rather than simply screens.

Frequently Asked Questions (FAQs)