Digital twin modeling: a vision of an AI future with UX at the helm

How UX can work with AI to create a more fulfilling future.

“UX in AI” has become one of the most confusing buzzwords in our industry.

Jakob Nielsen has famously talked about how UX is desperately needed for AI, but few can define what this means (or how to do it).

• Is it about designing chat interfaces and chatbots?
• Is it about working with algorithms or vibe coding?
• Is it about using Replit and Bolt instead of Figma?

The confusion is understandable.

When people say “UX in AI,” they’re usually describing a collection of loosely related activities without a unifying framework.

However, there is an existing AI use case that not only cuts through ambiguity but also demonstrates a clear path for UX designers to achieve these skills: digital twin modeling.

Why “UX in AI” Feels So Vague

“UX in AI” isn’t an unimportant topic, but there might be a dozen different interpretations for what that means, such as:

• Designing interfaces for AI tools (ChatGPT’s chat bubbles, Midjourney’s prompt boxes)
• Doing “vibe coding” to turn mockups into fully developed applications
• Making AI outputs more explainable (translating outputs, adding explanations, and sources)
• Creating knowledge-based repositories (Second Brains/RAGs)
• Creating AI-powered personalization (Netflix recommendations, Spotify playlists)
• Building conversational experiences (chatbots, voice assistants)
• etc.

Each one of these might be a valid instance of UX in AI, but lumping them together makes a field so broad it’s impossible to master.

In fact, many designers I’ve talked to wonder, “How do I even start? What skills do I need? What are companies looking for?”

Understanding digital twins helps you answer those questions.

How Digital Twins Saved Apollo 13 (and NASA)

The term “digital twin” comes from NASA, born from one critical question: How can we monitor and maintain things that are hard to access?

When Apollo 13 suffered its famous explosion (which coined the phrase “Houston, we have a problem”), NASA engineers worked around the clock with a replica spacecraft on the ground, which was the first digital twin.

This physical replica allowed them to test solutions safely before transmitting instructions to the endangered crew 200,000 miles away.

Modern digital twins have evolved far beyond NASA’s original concept, extending into industries where monitoring and optimization are critical:

• Wind turbine motors mounted 100 feet above ground, where sending a technician for routine checks is expensive and dangerous
• Smart buildings managing hundreds of rooms, each with different occupancy patterns and energy needs
• Medical equipment networks across hospital floors, predicting failures before they impact patient care
• GE’s jet engines embedded with thousands of sensors, monitored continuously without grounding planes

It’s in these cases, though, that UX becomes critical to answer one question:

• What is the essential data that users need to make decisions?

The answer determines whether your digital twin becomes a powerful decision-making tool or an overwhelming data dump.

This also highlights the role of UX in AI, specifically its ability to predict and influence user behavior.

Using AI to Predict and Change User Behavior

Imagine having a Healthcare Digital Twin designed to ensure people take their medication every day. The simplest version might be an alarm that goes off every day at the same time.

However, if you’re working with AI, you might be able to build a system that models each patient’s adherence patterns (i.e., whether they’re likely to take medication every day).

For example, what if Greg woke up 45 minutes late today, or if he skipped breakfast? He might be much more likely to skip taking his medication.

From there, you can predict when they are likely to skip doses and automatically adjust reminder timing, tone, and frequency to maximize the likelihood of taking medication as directed.

This is the power of a digital twin model. Rather than simply designing a simple UI that provides daily reminders, digital twins rely on real-time learning and automatic adaptation to improve based on actual user behavior.

Digital Twin-Powered UX Process: Research → Design → Build → Launch → Continuous real-time learning → Automatic adaptation

Doing this helps address three critical problems designers often encounter.

The Three Design Problems Many Designers Face

There are three challenges that designers currently face that digital twins can help address.

The “Design in the Dark” Problem

At its core, many design decisions are based on outdated analytics, limited research, and educated guesses. You launch features and wait months to see if they work.

However, when your team wants to build a digital twin, you get continuous visibility into how the design performs in actual use. Rather than struggling to find time for user research, you gain continuous insight into user patterns, struggles, and opportunities.

The “Set It and Forget It” Trap

Unfortunately, many designers’ work ends at launch, which leaves you disconnected from real-world performance. Users adapt, struggle, or abandon experiences as you move on to the next project.

However, when you design systems that learn and adapt automatically, your experiences become more effective over time, not less, because they’re continuously optimizing based on user behavior patterns.

The “AI Skills Gap” Challenge

The industry is moving toward data-driven, AI-assisted experiences, but most designers lack the knowledge to participate meaningfully in this shift.

However, learning to create digital twins provides a concrete framework for applying AI to UX problems. Instead of learning “AI” in the abstract, you learn to build behavioral models that solve specific user experience challenges.

Sounds Great, But How Do I Learn About Digital Twins?

Here’s the downside to digital twins: you can’t jump straight to building them. They’re essentially advanced data-informed design on autopilot.

But before you can build systems that automatically interpret and act on user data, you need to master the fundamentals:

• Metric Selection: Which user behaviors actually predict experience quality?
• Experiment Design: How do you create and test hypotheses that yield actionable insights?
• Data Translation: How do you turn behavioral patterns into design improvements?
• Continuous Optimization: How do you build measurement and adaptation into your design process?

However, the learning path is significantly more straightforward for this use case than for other AI paths.

Greg Nudelman, in his book UX for AI, says most AI projects (and learning) will fail for one particular reason: they fail to identify the correct use case.

The AI landscape is no different right now. Learning Replit might make you employable, or it might be a useless skill in 3 years.

Chatbots might fall out of favor, or vibe coding might be seen as a cheesy fad in a few months.

But digital twin modeling? That’s a solid use case, and your learning starts with one specific question:

• What user behaviors predict success with an experience?

If you’re able to start by isolating specific user behaviors to seek out and design for, you’ll be on your way toward learning how to build a digital twin.

Here’s a Basic Timeline of How to Learn Digital Twin Modeling:

Phase 1: Master Data-Informed Design:

Learn to identify how user behavior impacts metrics. Design experiments to change user behavior (and business outcomes), and translate user findings into actionable business insights.

Phase 2: Study Digital Twin Applications:

Understand what makes digital twins different from analytics dashboards. Understand how Netflix creates behavioral models of viewer preferences or how Tesla uses driving pattern data to improve autonomous features.

Phase 3: Build Behavioral Models:

Start small: Create simple systems that identify poor user behavior patterns and adapt around them. A form that adjusts based on completion rates. A navigation that reorganizes based on usage patterns.

Phase 4: Scale to Intelligence:

Build systems that don’t just adapt, they predict. Experiences that prevent problems before users encounter them.

The Bottom Line: Clarity Through Modeling

Currently, “UX in AI” is a topic that many in the design community are grappling with.

Whether it’s about learning the right tools or convincing your team not to take specific approaches, there doesn’t seem to be a clear path toward learning AI.

However, digital twin modeling offers a clear and practical path forward for many UX professionals to understand the impact they can have with AI-powered products.

Instead of wondering how to apply AI to design, you focus on building real-time models of user behavior that enable better experiences.

If you’re interested in learning how to apply AI to UX but aren’t quite sure how it’s useful, consider learning more about digital twin modeling.

It’s a tried and true use case that hasn’t just saved projects: it’s saved lives.

Want to learn the first phase of digital twin modeling? I teach designers how to do this in 4 weeks.

Kai Wong is a Senior Product Designer and Data and Design newsletter author. He teaches a course, Data Informed Design: How to Show The Strategic Impact of Design Work, which helps designers communicate their value and get buy-in for ideas.

Digital twin modeling: a vision of an AI future with UX at the helm was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.

This post first appeared on Read More