Co-Design is the Future

Every few years, someone declares that design thinking is dead. The Double Diamond is too linear. Design sprints are theatre. The Post-it note has become a symbol of everything wrong with how organisations pretend to innovate. The critique lands, gets shared widely among designers who feel vaguely guilty about their own practice, and then — nothing much changes, because the critique rarely comes with anything better.

The latest version of this cycle is the Stingray Model, which proposes a more dynamic, iterative alternative to the Double Diamond’s two-phase diverge-and-converge structure. It’s a thoughtful piece of work. It reflects genuinely important observations about how design actually unfolds in complex organisations, where problems don’t wait politely to be fully defined before solutions start emerging, and where the neat separation between research and ideation is mostly a fiction maintained for the benefit of project plans.

But I don’t think the Double Diamond is the problem. And I don’t think replacing it with a better diagram is the solution.

The problem is who’s in the room.

What the Double Diamond actually says

The Double Diamond, developed by the British Design Council in 2005, describes a process of two divergent-convergent phases: first you expand your understanding of the problem, then you converge on a problem definition; then you expand your exploration of solutions, then you converge on something to build. It is deliberately simple. It was designed to give non-designers a legible mental model for why design takes time and doesn’t go in a straight line from brief to output.

The critics are right that this simplicity is also a limitation. Real design work is messier. Problem and solution co-evolve. A prototype reveals something about the problem that changes the brief. Research done in the solution phase uncovers needs that were invisible in the research phase. The diagram doesn’t capture any of this.

But the diagram was never meant to capture all of this. It was meant to communicate one important thing: that you need to understand before you decide, and that understanding requires expanding before you can usefully collapse. That insight is not wrong. It has not been superseded by AI. It is, if anything, more important now than it was in 2005, because the pressure to skip directly to solution has never been higher.

When anyone can generate a plausible-looking interface in thirty seconds with a prompt, the temptation to treat that output as a starting point — rather than as a very confident answer to a question nobody has properly asked yet — is significant. The Double Diamond, whatever its limitations as a process map, is a useful reminder that the question matters as much as the answer.

---

What’s actually broken

What is broken about how design thinking gets practised is not the framework. It’s the relationship between the designers and the people the design is for.

The canonical design thinking process treats the people who will be affected by a design as subjects of research rather than participants in the work. You go out, you observe, you interview, you synthesise, you return to your studio and you make sense of what you found. You generate ideas. You prototype. You test. You iterate. The people you spoke to in the research phase may or may not see the thing that gets built.

This model has produced real value in contexts where the people doing the designing have genuine expertise and the people being designed for lack the tools to articulate what they need. A skilled interaction designer watching someone navigate a confusing form can see things the user cannot name. That observational expertise is real and worth preserving.

But it has also produced, consistently and predictably, designs that are optimised for the assumptions of the designers rather than the reality of the people using them. Government services that work beautifully for people who are digitally confident, native English speakers, in stable housing, with smartphones. Health apps designed by people who don’t live with the conditions the apps are meant to support. Financial tools built by people who have never experienced financial precarity.

The failure mode is not a methodology problem. It is a proximity problem. The people who most need good design are the people furthest from the rooms where design decisions get made.

Co-design is not a technique. It’s a commitment.

Participatory design — designing alongside the people who will be affected by the design, rather than on their behalf — has a longer history than design thinking. It emerged from Scandinavian labour movements in the 1970s, where workers fought for the right to be involved in decisions about the technology that would change their jobs. The original insight was political before it was methodological: the people whose lives are affected by a design have a legitimate claim to participate in making it.

That framing has been largely lost in the way co-design is currently practised. It has been absorbed into the design thinking toolkit as a technique — a workshop format, a way of generating ideas with users, a box to check on the empathy section of your project plan. Run a co-design session. Gather insights. Return to your studio. Produce a deliverable.

This is not co-design. This is participatory research with better branding.

Genuine co-design means that the people you are designing with have meaningful influence over the outcome. It means that their knowledge — about their own lives, their own contexts, the specific textures of the problem you’re trying to solve — is treated as expertise, not just as data. It means that the process is designed so that they can actually participate: sessions happen at times and in places they can reach, materials are accessible, the pace allows for genuine engagement rather than just the extraction of insights. And it means that there is accountability for what happens to their input — that people can see how their participation shaped what got built, or understand honestly why it didn’t.

This is harder than doing research. It takes longer. It produces more complexity in the short term, because when you design with people whose lives are different from your own, you encounter constraints and priorities and ways of seeing that don’t fit your initial mental models. That complexity is not a problem to be managed. It is the information.

---

Why AI makes this more urgent, not less

The arrival of AI into the design process has changed many things. It has made certain kinds of generative work faster and cheaper. It has lowered the barrier to producing high-fidelity prototypes early in the process. It has made it possible for designers to explore more options, more quickly, than was previously practical. These are genuine capabilities and they are worth taking seriously.

What AI has not done — what it cannot do — is replace human judgment about what is worth building. It cannot tell you who is being left out of your mental model of the user. It cannot notice the things that don’t fit the pattern. It cannot feel the friction of an experience that is technically functional but humanly degrading.

There is a particular failure mode I’m watching develop in real time, in design teams that are adopting AI tools quickly without changing their underlying process. The speed of generation has increased dramatically. The time spent in direct contact with the people being designed for has not increased proportionally. In some cases it has decreased, because the pressure to show outputs early is even higher when outputs can be produced quickly. The ratio of making to understanding is moving in the wrong direction.

AI systems, trained on data that reflects existing distributions of representation, are not neutral on this question. They amplify the assumptions embedded in their training data.

A design process that uses AI generation without the corrective pressure of direct engagement with real people in real contexts will produce, with greater speed and confidence, designs that serve the well-represented at the expense of the poorly-represented.

Co-design is not a corrective to AI. It is the practice that gives AI a good problem to work on.

---

What this means for how we work

I am not arguing for the abolition of Double Diamonds or design sprints or any of the other frameworks that get periodically declared dead. These are scaffolding. Scaffolding is useful. What you’re building inside it is what matters.

What I am arguing is that the question of who is in the room — who participates in defining the problem, who generates ideas, who evaluates prototypes, who has standing to say whether a design is working — is not a question that any methodology answers on its own. It is a question about power and proximity. It is a question that has to be answered before the methodology question, because it determines what the methodology is actually for.

The Double Diamond is not dead. The Stingray is not replacing it. What’s happening, or what needs to happen, is something more fundamental: a shift in the basic relationship between designers and the people they design for. From subjects to participants. From data to collaborators. From research to partnership.

AI will generate options faster than any human designer. That speed is real and it’s not going away. The question of which options are worth generating, which problems are worth solving, and which people the solutions actually need to work for — that question requires the presence of the people the design is for. Not as an input. As a voice.

The room needs to be bigger. That’s the methodology update that matters.