For years, the Double Diamond has given designers a simple way to explain how we work:

Discover, Define, Develop, Deliver

We explore the problem, narrow our understanding, explore possible solutions, and eventually converge on something worth building. It is neat and understandable, and if we are being realistic, rarely as linear as the diagram suggests. But AI is introducing a more fundamental change.

The interesting question is no longer:

How can designers use AI to move through the Double Diamond faster?

I think there is a bigger question:

What happens to the Double Diamond when designers are no longer the only people doing design activities?

That shift may ultimately be more important than the productivity gains AI gives us.

1. Creation has become democratized

A few years ago, many activities in the product development process required specialist skills. A designer might conduct research, synthesize findings, map journeys, explore concepts, create user flows, write interface copy, produce wireframes and eventually build a prototype. Today, AI can assist with almost every one of those activities.

Give an AI system a collection of research transcripts and it can identify recurring themes. Describe a product problem and it can generate several possible user flows. Give it requirements and it can produce interface concepts. Ask it to critique a journey and it can identify potential usability issues.

And increasingly, these capabilities are not available only to designers.

  • Product managers can prototype.
  • Engineers can generate interfaces.
  • Founders can explore product concepts.
  • Researchers can create interaction ideas.

This changes something fundamental. The ability to produce a design artifact is becoming less scarce, and when production becomes less scarce, the value of design starts moving somewhere else.

2. The Double Diamond starts to compress

The traditional Double Diamond contains two major divergent phases. First, we explore the problem space. Then, we explore the solution space. Historically, both required considerable human effort. Imagine a team exploring five potential solutions: designers might sketch different approaches, create wireframes, prototype promising directions and gradually narrow the options, a process that could take days or weeks. AI can generate dozens of plausible directions in minutes. That doesn’t mean those directions are good. It means generating possibilities has become dramatically cheaper, and that distinction matters. AI doesn’t necessarily remove divergence from the Double Diamond. It accelerates divergence.

Instead of asking:

Can we generate another solution?

the harder question becomes:

Which of these solutions deserves our attention?

The bottleneck begins moving from generation to judgment.

3. Designers become evaluators, not just creators

This is where I think the designer’s role becomes more interesting. Imagine an AI system generates ten potential experiences for a new product. The difficult part isn’t producing the eleventh. The difficult part is understanding:

  • Which one solves the actual customer problem?
  • Which interaction fits the user’s mental model?
  • Which introduces unnecessary cognitive load?
  • Which aligns with the business model?
  • Which can realistically be built?
  • Which creates unintended behaviour?
  • Which should never be built at all?

These are not primarily interface-production questions. They are judgment questions, and judgment depends on understanding users, systems, technology, business constraints and human behaviour.

This changes what design expertise looks like. The designer becomes less of the person responsible for producing every solution and more of the person responsible for helping the organisation recognise the right solution.

4. AI introduces a third design problem: behaviour

Traditional digital products were largely deterministic. A designer could specify:

User clicks X, system does Y.

AI products don’t always behave like that. The same input can produce different outputs. The system may misunderstand intent, provide incomplete information, take an action the user didn’t expect, or sound confident while remaining uncertain. And increasingly, AI systems don’t just respond to users. They may act on their behalf.

That creates an entirely different category of design questions.

  • How much autonomy should the AI have?
  • When should it ask permission?
  • When should it explain its reasoning?
  • How should uncertainty be communicated?
  • How can users correct the system?
  • What happens when AI gets something wrong?
  • When should a human take over?

And perhaps most importantly:

How do we make an intelligent system feel predictable enough for humans to trust?

This is where I believe designers become increasingly important. We aren’t only designing interfaces anymore. We are designing relationships between humans and intelligent systems.

5. Human-in-the-loop becomes a design principle

One of the most important concepts in AI product design is the human-in-the-loop. AI can generate, recommend, predict and increasingly execute, but capability does not automatically mean autonomy.

Consider an AI assistant helping someone manage payments. It might be reasonable for AI to categorise transactions, identify unusual activity, suggest payment dates, or prepare a transaction. But should it automatically move £20,000 between accounts? Probably not without additional controls.

The design question therefore becomes:

Where should AI act, and where should humans retain control?

This creates several layers of interaction:

  • AI proposes, human decides
  • AI executes, human confirms
  • AI acts, human monitors
  • AI fails, human intervenes

These aren’t edge cases. They are becoming fundamental interaction patterns for AI products. Designers therefore need to think beyond screens and flows toward levels of autonomy, trust and control.

6. The Double Diamond becomes more collaborative

There is another consequence of AI that I think receives less attention. AI doesn’t just accelerate designers. It gives other disciplines access to design capabilities. A PM can arrive at a discussion with a working prototype. An engineer can generate an alternative interaction. A researcher can turn insights into interface concepts.

That could feel threatening if we define design as ownership of design artifacts. I see it differently. I believe it can produce better product decisions, because the goal was never for designers to own Figma. The goal was to build better products.

If more people can express ideas visually and test assumptions earlier, the product development process becomes more collaborative. The designer’s responsibility then shifts from:

I create the solution.

toward:

I help the team understand which solution should exist and why.

That is a much more strategic role.

7. A possible AI-era Double Diamond

So what might the design process actually look like? I don’t think we need to throw away the Double Diamond. The fundamental principle (diverge, understand, converge, validate) still makes sense, but the mechanics underneath it are changing.

A modern AI-assisted process might look something like this:

  • Problem. Start from a real product or user tension.
  • Human + AI discovery. Research, behavioural data, market signals, support conversations and existing product knowledge.
  • AI-assisted synthesis. Themes, hypotheses, opportunity areas and contradictions emerge faster.
  • Problem framing. Humans determine what problem actually deserves solving.
  • AI generates possibilities. Flows, concepts, prototypes, content and interaction models can be explored rapidly.
  • Cross-functional exploration. Designers, PMs, engineers, researchers and domain experts evaluate possibilities together.
  • Design judgment. Solutions are evaluated against user needs, business outcomes, technical constraints, accessibility, trust and risk.
  • User validation. Real users expose where our assumptions, and AI’s assumptions, are wrong.
  • AI-assisted refinement. Solutions evolve rapidly based on evidence.
  • Continuous learning. Product behaviour informs the next cycle.

The Double Diamond hasn’t disappeared. It has compressed and become more continuous.

8. The skills designers need are changing

If AI makes execution faster, designers shouldn’t compete with AI on execution speed. The differentiating skills move higher up the value chain.

Problem framing

AI is excellent at generating answers, but the quality of those answers depends heavily on whether we’re solving the right problem. Understanding which problem deserves attention becomes increasingly valuable.

Systems thinking

AI experiences rarely exist as isolated screens. They involve data, models, permissions, automation, human intervention and downstream consequences. Designers need to understand the system surrounding the interface.

Product judgment

When AI can generate twenty plausible solutions, someone needs to recognise which direction creates meaningful value.

Designing for uncertainty

Traditional interfaces often communicate certainty. AI systems operate in probabilities. Designers need patterns for communicating confidence, uncertainty, limitations and failure.

Business understanding

A beautiful experience that doesn’t create sustainable business value isn’t a successful product. Senior designers increasingly need to connect:

User outcomes to product behaviour to business outcomes.

Responsible AI judgment

Bias, explainability, transparency, privacy and autonomy cannot simply be handed to engineering or legal teams after the product has been designed. They are design decisions.

9. The designer becomes an architect of intelligent systems

For a long time, digital product design has been closely associated with interfaces: screens, components, flows and prototypes. Those things aren’t disappearing, but I think they will become a smaller part of what defines strong product designers. The emerging role looks broader.

Designers become:

  • Problem framers who determine what deserves solving.
  • Pattern experts who understand how humans interpret complex systems.
  • Decision architects who structure how people make choices.
  • Human-in-the-loop designers who determine where humans and AI share control.
  • AI behaviour designers who shape how intelligent systems respond, explain themselves and recover from failure.

And perhaps most importantly:

Evaluators of possibilities.

Because in a world where AI can generate almost unlimited possibilities, the scarce resource isn’t creation. It’s judgment.

The Double Diamond isn’t dead

AI isn’t removing design from product development. It’s changing where design expertise creates value.

The first era of digital product design rewarded our ability to create interfaces. The next may reward our ability to understand systems, frame problems, evaluate possibilities and design responsible relationships between humans and intelligent machines.

So I don’t think the Double Diamond disappears. I think it compresses. Creation becomes faster, participation becomes broader, and iteration becomes continuous. And the role of the designer moves from being primarily the creator of interfaces toward becoming an architect and governor of intelligent systems.