How Do You Evolve a Design Organization for an AI-Native Future?
My company announced we were becoming AI-native. Effective immediately.
Which meant our design organization needed to evolve. I needed some practical guidance for someone responsible for transforming a design organization. Fast.
Normally, I’d start the way I always do. Read what smart people have written. Look at how other companies are approaching the problem. Borrow a few good ideas. Form a point of view.
This time was different.
There were plenty of articles about using AI tools to generate wireframes faster, the joy of vibe coding, and predictions about AI replacing designers now that everyone and their mother can build a prototype.
What value does Design bring? What capabilities should we be building? What becomes more important? What entirely new responsibilities emerge?
So I tried a skills gap analysis. I’d figure out what we have today and what we’ll need for tomorrow. I asked Claude. I asked ChatGPT. I assembled a long list of skills and capabilities. Impressive sounding. I didn’t even understand many of them.
AI was generating answers faster than I could generate the right questions.
Turns out I couldn’t identify the skills we’d need without first understanding the work we’d be doing. So I took a step back and asked: How will our work change? What will remain the same?
The big shift is at the core of the experience. Traditional software is largely deterministic. We decide how it behaves. Users interact with it. In contrast, intelligent systems can recommend, adapt, reason, delegate, and sometimes act on a person’s behalf.
Traditional software required us to define behavior. Intelligent systems require us to define boundaries, relationships, and responsibility.
Organizations suddenly need to answer questions that didn’t exist before.
- How much autonomy should a system have?
- How should humans and AI work together?
- When should a system act versus ask?
- How do people understand and trust recommendations they didn’t explicitly request?
- How do we evaluate behavior that isn’t fully scripted?
At the same time, some familiar questions still apply — are we solving the right problem, where are people struggling, what creates value, what risks are we introducing? The difference is now they share space with the harder set.
The more I explored, the more I realized I wasn’t ready to talk skills. I was looking for the questions organizations would need to answer as products become more intelligent. Once I understood those questions, I could start thinking about the capabilities required to answer them.
What emerged was a framework I now use to think about the design of intelligent systems.
This framework describes five dimensions that organizations must consider when creating intelligent systems. They apply across a spectrum of intelligent systems, from AI-assisted experiences to autonomous and agentic systems.
The framework is not a process.
Get Deanna Aho’s stories in your inbox
Join Medium for free to get updates from this writer.
It’s not a maturity model.
And it’s not something Design owns.
Instead, it describes critical questions that organizations must answer. This framework highlights Design’s perspective and contribution.
The five dimensions are:
Problem Framing — What problem is actually worth solving and can AI genuinely help?
Defining Intent — How should the system behave?
Designing Human-AI Collaboration — How should humans and AI work together?
Learning Through Experimentation — What do we need to learn before moving forward?
Moving Through Ambiguity — How do we make decisions when we don’t know enough?
Together, these dimensions describe the questions organizations must answer when creating intelligent systems, regardless of the technologies, tools, or models involved.
One of the things I like most about the framework is that it shifted my thinking about Design’s role. The future of Design isn’t simply learning new AI tools. It’s helping organizations answer increasingly complex questions about intelligence, behavior, trust, collaboration, learning, and decision-making.
That realization led to my next challenge.
If these are the questions organizations need to answer, what capabilities do we need to answer them? That exercise led to a second model that groups capabilities into three categories:
Core capabilities that remain foundational.
Expanding capabilities that become more important in AI environments.
Emerging capabilities that many design organizations do not yet consistently possess.
Looking at the two artifacts side by side, it was clear they weren’t separate exercises. The capabilities naturally mapped back to the framework dimensions. That’s when I added the capability row.
The specific capabilities will evolve. The tools will evolve. The technology certainly will. The questions themselves will endure.
When I started this exercise, I thought I was trying to identify the skills our team would need in an AI-native future. What I discovered was that skills were the wrong starting point.
Before I could determine what capabilities to build, I needed to understand the questions we’d be responsible for helping answer.
And that’s ultimately what this framework gave me.
Not a prediction about the future.
A way to make decisions despite the uncertainty.
A way to decide what to build, what to learn, and where to invest.
