AI Has Collapsed the Double Diamond
What happens to product design when the blank slate disappears and simulation replaces observation?
For years, the double diamond was the gold standard for product design thinking. It gave teams a shared roadmap:
- Diverge to understand the problem
- Converge to define it
- Diverge again to explore solutions
- Then converge toward implementation
In a world defined by ambiguity and constraint, it offered structure without rigidity—sequencing without suffocation.
But that clarity no longer holds.
Generative AI doesn’t respect process.
What once unfolded over weeks or months now happens in hours — sometimes minutes.
You don’t begin with a research plan or a blank page. You begin with a prototype. A conversation. A set of plausible futures generated by a model trained on millions of pasts.
We’re witnessing process collapse — not as failure, but as compression. Discovery and delivery no longer live in sequence. They co-occur, overlap, and in some cases, reverse.
At the same time, the designer, the PM, the strategist — they’re shifting. From creators to curators. From inventors to interpreters. In the age of AI, the hardest part isn’t starting. It’s choosing.
The Double Diamond Was a Good Idea
The double diamond reflected a deeper truth: innovation is not linear.
To design something useful, you need to wander — empathize, observe, frame. Then you narrow in. The second diamond mirrored this rhythm for solutions: generate many ideas, test, refine, converge.
It created clarity between roles. Researchers owned discovery. Designers ideated. Engineers delivered. Stakeholders could trust in the logic of progress.
But that same structure baked in latency. Ideas moved slowly because certainty had to be earned.
Process Collapse: When Discovery and Delivery Merge
Generative AI collapses the distance between idea and execution. A prompt becomes a prototype. A sketch becomes a simulation.
What once belonged to the “build” phase is now how we think.
I recently sat in on a roadmap review where the engineering team had already shipped a working demo — before product had finalized the problem statement.
It wasn’t an outlier. It was a signal.
The team wasn’t reckless. They were working with tools that collapse the “how” before the “why” is even resolved. Tools that make exploration and execution indistinguishable.
Design sprints that used to take five days now compress into five hours. GPT writes product briefs. Midjourney generates 60 design directions in seconds. Copilots turn back-of-the-napkin ideas into functioning features.
Exploration happens through implementation.
We used to define the problem, then solve it. Now we test the solution to figure out what problem it might solve.
The End of the Blank Slate
This shift begins even earlier.
You no longer start from zero.
At a design offsite last quarter, a colleague prompted Midjourney to generate 60 brand directions for a wellness app — in under an hour. She didn’t choose one. She built a moodboard from 20. Her work wasn’t creation. It was orchestration.
Her creativity didn’t start on a blank canvas. It started in a flood. Her skill was discernment — knowing what to pursue and what to let go.
This is the new normal.
The blank page is gone. In its place: an infinite scroll of maybes.
We used to measure creative success by the spark of a new idea. Now, it’s our ability to sift signal from noise. Curation becomes the new craftsmanship.
The Risk of Premature Convergence
There’s no question: this shift accelerates insight.
But speed invites risk — especially the risk of falling in love with early outputs.
A plausible prototype can create false confidence. When you can build before you understand, you may forget to understand at all.
The system starts to optimize for what’s easy to build — not necessarily what matters.
AI doesn’t just collapse process — it compresses ambiguity. And while ambiguity is uncomfortable, it’s where insight lives.
This isn’t a case for slowing down. It’s a case for adding intentional friction — for pausing inside the rush.
When anything can be built instantly, the discipline isn’t in shipping. It’s in stopping to ask, “Should we?”
Synthetic Empathy: Designing for Users We Simulate
Another shift is more subtle — but more profound.
We’re beginning to replace real users with simulated ones.
One team I work with runs LLM-based synthetic interviews — complete with fake transcripts and sentiment summaries — before a single usability session is run.
It’s efficient. But it raises a question: are we listening to users, or to ourselves through a predictive filter?
We now model behavior, simulate feedback, and pressure-test interfaces using machine-trained stand-ins. These synthetic signals are fast. Scalable. Convincing.
But they shift the foundation of empathy from experience to inference.
AI lets us simulate the market before we’ve spoken to a single customer. But simulation isn’t empathy — it’s a mirror trained on the past.
Synthetic empathy helps us move faster, but it can also flatten human nuance. Patterns are not people. Predictions are not needs. Without care, we may begin designing for what’s legible, not what’s lived.
New Capabilities for a Post-Process World
We don’t need to mourn the double diamond. But we do need to move past it.
A new mode of practice is emerging. And with it, new capabilities:
- Curatorial Thinking — discernment as creative leadership.
- Promptcraft as Strategy — designing inputs to probe possibility.
- Synthetic Research Fluency — merging inference and interaction.
- Real-Time Co-Creation — working with AI in feedback loops, not handoffs.
This isn’t a rejection of process — it’s a reformation of it. Less of a pipeline, more of a pattern. Less sequencing, more simultaneity.
The New Beginning
AI hasn’t killed the double diamond.
It’s just made it irrelevant as a sequential guide.
The new challenge isn’t how to preserve our process. It’s how to navigate abundance. To lead when the page is already filled. To teach teams not how to generate , but how to judge.
Not just how to move fast — but how to move wisely.
Because in the age of AI, the hardest part of building isn’t knowing what to do next.
It’s knowing where to begin.
