Design Thinking for AI: A 5-Stage Framework Every Builder Needs
We're using 20th-century design methods for 21st-century AI problems. Here's what Design Thinking might look like in the age of AI.
Design Thinking has had a rough decade. It went from breakthrough to buzzword to meme. Natasha Jen (incidentally I am a big fan of her work) called it “post-it note theatre” (and also “bullshit”), Don Norman said it was “a useful myth.”
Fair points, honestly.
I’ve been one of those people too, and one of those who quietly dropped it, got tired of the stickies and stopped bringing it up.
Recently though, I’ve found myself circling back to it. I’ve been playing around with adapting it to this AI Era. Also I really like a good framework.
Why now?
AI has changed the texture of design. We’re not designing for people alone anymore, we’re designing with/for intelligence. That changes everything. I’m obviously not the only person thinking about how Design frameworks evolve.
Recent research has started reframing what this looks like. Adam Fard calls it “AI-First Design Thinking.” Weisz, He, and Muller (2024) propose six design principles for generative AI that move beyond the empathy-prototype-test loop:
And Zhangqi Liu and colleagues (2025) describe a “co-creation spectrum,” mapping how humans and AI collaborate from passive assistance to proactive collaboration.
As an aside, this isn’t a “Defence of Design Thinking” post, it’s more like an experiment. A bit of a noodle on what it might look like if we reshaped it, and gave it new language for an AI era.
I’ll let you decide if it deserves a second life.
Here’s what Design Thinking looks like in the age of AI 👇
1. Empathize → Anticipate
Design once began with understanding human needs.
Now, it begins with anticipating how humans and AI will interact emotionally, ethically, and cognitively.
“Empathy” once meant listening to users to understand their needs. Anticipation means sensing what might emerge when intelligence meets intention. It’s about mapping not just user journeys, but human–AI relationships.
This echoes Zhangqi Liu’s (2025) notion of “anticipatory co-creation,” where designers predict shifts in human behaviour as AI takes a more proactive role.
Ask yourself:
Where will AI amplify human ability and where might it undermine it?
When does AI build confidence, and when does it trigger anxiety?
What happens to trust when the system surprises us?
When we design for AI, we’re necessarily designing for uncertainty. Empathy helps us connect, anticipation helps us prepare and navigate unpredictable outcomes .
Potential Tool: Map “AI anxiety points”, moments when users hesitate, doubt, or delegate too much.
2. Define → Decode
In traditional design, we define the problem, working hard to understand it well.
In AI design, we decode which problems are actually worth solving with AI (and which are not).
Design Thinking failed when it became a formula and we stopped doing the thinking. AI makes this even more urgent. Not every problem is an AI problem. Some are human, cultural, or systemic.
Mohme, Bick & Böckle (2024) highlight this tension in their paper Design Thinking and Human-Centered AI, arguing that ethical AI design requires decoding, in other words where artificial intelligence belongs and where it doesn’t.
Before building, pause and decode:
Is this problem better solved through intelligence, or through empathy?
Are we addressing the cause or just the symptom?
What’s lost when we hand this decision to a model?
Designers in the AI era must think less like product managers and more like philosophers and ethicists: defining not only what’s possible, but what’s permissible.
Potential Tool: Use a “Should-We-AI?” checklist before any build.
3. Ideate → Integrate
The ideation phase used to be about creativity.
Now, it’s about integration as well: finding the human+AI sweet spot.
Jon Kolko once argued that Design Thinking divorced design from its craft. I think that AI gives that craft back in the form of Generative Collaboration, where humans set the direction and AI expands the possible.
This aligns with Weisz et al. (2024), who note that designers must balance exploration and control when working with generative systems. We’re moving from AI as a passive assistant to a proactive design partner.
Ask yourself:
What’s the human genius here?
What’s the machine genius?
How do they co-create something neither could alone?
When humans and AI work in tandem and when prompts, sketches, and prototypes evolve together, we start designing systems that are meeting the potential afforded to us in this moment.
Potential Tool: Run “Human+AI Jam Sessions” and co-creation sprints where prompts and prototypes evolve in dialogue.
4. Prototype → Prompt
In classic design, prototypes help us test ideas.
In AI design, prompts become our prototypes. Each prompt is a small experiment, a conversation with intelligence. It reveals how a model interprets, reasons, and learns from context.
Prompting isn’t a technical skill alone, it’s also a design practice.
In some ways, iteration with generative models is a kind of dialogue that requires designers to externalise and refine their intent.
Potential Tool: Treat prompt logs as design artifacts : versioned, annotated, and shared.
5. Test → Trust
Traditional testing measures usability and performance.
AI testing measures trust. We can no longer rely on metrics like accuracy or efficiency alone. Trust is the new usability. Transparency, explainability, and ethical alignment are now part of the design brief.
Trust in some ways is the connective tissue between AI and human agency.
In this AI-era, Design Thinking can suggest new validation modes like participatory testing, interpretability mapping, and continuous alignment loops.
Ask:
Does this AI behave as intended and as expected?
Would I trust this system to make a decision on my behalf?
How do I know when to step in or step back?
Designers need to stop only prototyping experiences, but expend their mindset to prototyping relationships.
Potential Metric: Measure “Trust Delta”: the gap between what users expect AI to do and what it actually does.
Design Thinking has been crushed under the weight of its own hype. Maybe AI gives us a reason to bring back the good parts without the hype.
This exploration of it isn’t about workshops, stickies or templates. It’s about reframing how we think, collaborate, and build with intelligence.
So no, I’m not trying to resurrect Design Thinking as it was.
I’m just wondering aloud what it could become and what a better framework to work with AI looks like.
Here’s a start, anyway.
Great thoughts MC. I just walked out of a conversation with a few parents and teachers at Jonathan and April’s school just now looking at the role of tech and AI.
My brain is now trying to think through frameworks and guardrails that can help children develop healthy collaboration techniques to problem solve (design thinking), while also blending in AI and general digital usage.
Trying to think about developing those healthy and productive techniques at an early stage of a child’s development is interesting. A little bit tangential but some of your thoughts above are super helpful to my thinking.