AI is a tool, not a designer

Design is not output. It's decision-making. The real work in design happens long before anything is created. It sits in understanding the problem properly, interpreting user behaviourglossary×User BehaviourUser behaviour refers to how users interact with a product, including actions, patterns, and decision-making processes.Open glossary term→, navigating constraintsglossary×ConstraintsConstraints are limitations or restrictions that impact how a product or solution can be designed or built.Open glossary term→, and making trade-offsglossary×Trade-offsTrade-offs are decisions where improving one thing requires giving up another, such as speed against quality or scope against cost. Every product decision involves them, whether or not they are made explicit.Open glossary term→ that balance business needs with real-world usage. None of that is visible in the final screens. But all of it determines whether those screens actually work.

The confusion is understandable, because output is the only visible part. Nobody sees the twelve journey variantsglossary×VariantA variant is a version of a design or experience used in testing or experimentation.Open glossary term→ that were ruled out, or the constraint that killed the obvious solution, or the two weeks spent establishing what the problem actually was. They see the screens. So when something else can produce screens, it looks like the job.

It is the same mistake as judging a codebase by how quickly it was typed. Speed of production and quality of decision are different measures, and only one of them is visible in a demo.

AI doesn't understand why something exists. It doesn't challenge assumptions. It doesn't question whether the problem has been framed correctly. It doesn't weigh up competing priorities or consider the long-term impact of a decision. What it does is generate. And it does that well. It works from patternsglossary×PatternA pattern is a reusable solution to a design problem that recurs across products and contexts. It captures an approach already shown to work, so a team solves the problem once rather than every time it appears.Open glossary term→: taking existing structures, recognising common approaches, and producing something that looks right based on what it has seen before. But it's working from probability, not understanding. Something that looks right is not the same as something that is right.

This shows most clearly when the right answer is to buildglossary×BuildA build is the process of compiling and packaging code into a runnable application.Open glossary term→ nothing at all. A modelglossary×ModelA model is a system or representation used to process data and generate outputs, often trained to perform specific tasks.Open glossary term→ asked to design a booking flow will design a booking flow. It has no mechanism for coming back and saying the real problem is that people are ringing up because the pricing page is unclear, and the flow should not exist.

I've seen AI-generated outputs that are visually convincing, structurally sound, and technically coherent, but completely disconnected from the contextglossary×ContextContext is the set of surrounding conditions that shape how someone behaves and decides, including their goal, their environment, their time pressure, and what happened immediately before. The same action can mean different things in different contexts.Open glossary term→ they're supposed to operate in. Journeys that don't align with how the business actually works. Content that sounds polished but misses the nuance that buildsglossary×BuildA build is the process of compiling and packaging code into a runnable application.Open glossary term→ trustglossary×TrustUser confidence that a product, service, or organisation will do what it promises.Open glossary term→. Flows that look complete but fall apart under real usage. Everything appears finished, but nothing has been properly thought through.

What makes this expensive is that the failure stays invisible until late. A badly drawn wireframe gets challenged in a review because it looks unfinished. A beautifully rendered screen encoding the wrong assumption gets approved, built, and only questioned once real users reach it.

Design isn't about producing the artefact. It's about shaping the thinking behind it. Understanding what matters, what doesn't, what needs to be simplified, what needs to be prioritised, and what needs to be challenged: those decisions come from experience, from researchservice×User ResearchUnderstand user behaviour, validate ideas, and make clearer product decisions with evidence you can act on.Open service→, from contextglossary×ContextContext is the set of surrounding conditions that shape how someone behaves and decides, including their goal, their environment, their time pressure, and what happened immediately before. The same action can mean different things in different contexts.Open glossary term→, and from a clear understanding of how everything fits together.

Used properly, AI is incredibly effective at removing frictionglossary×FrictionFriction refers to anything that slows users down or makes it harder for them to complete a task. It can be caused by poor design, unnecessary steps, unclear messaging, or technical issues.Open glossary term→ from parts of the processglossary×ProcessA process is a defined sequence of steps used to achieve a specific outcome.Open glossary term→ that don't require deep thinking: drafting content, exploring variations, expanding on ideas, handling repetitive work. In those areas, speed is valuable. But the strongest use of AI sits alongside design, not in place of it. It supports the work but doesn't define it. AI is not the designer. It is the assistant. Its effectiveness depends entirely on how it is used. Used without direction, it produces more noise; used with clear thinking, it becomes a powerful way to move faster without losing quality.