Design matters in the Age of AI

Welcome to the new paradigm.

The history of technology is not a smooth, continuous progression from axe to AI. Rather, it is punctuated by paradigm shifts: moments when new capabilities arrive, changing what can be achieved, how people interact with the technology, and what kinds of value become possible.

These shifts dramatically impact culture and organisational behaviour. As new technologies such as personal computing, the internet and the smartphone arrived over the past century, they disrupted the corporate landscape, changed the behaviours and expectations of customers, redefined the kinds of companies that succeed, and transformed the disciplines required to translate raw capability into meaningful products and services.

Artificial intelligence represents one of these shifts.

The defining characteristics of this new paradigm are already becoming clear. Interaction is increasingly mediated through natural language rather than structured interfaces. Systems generate responses, rather than selecting from predefined options. AIs can interpret intent, synthesise information, operate across sequences of actions, and draw on tools and external systems. In some cases, they can plan, evaluate, use memory, and act with a degree of autonomy. This changes the nature of software itself.

Traditional software is largely deterministic. Given the same input, it tends to produce the same output. Its behaviour is specified in advance through rules, logic, workflows, and interface states. AI systems, by contrast, are probabilistic and semi-autonomous. Their behaviour emerges from models, prompts, retrieved knowledge, tools, orchestration logic, context, permissions, evaluation, and user interaction. AI systems do not simply execute instructions. They participate in reasoning-like work, judgement-like work, and action.

This introduces extraordinary opportunity. AI systems can support tasks that were previously difficult or impossible to automate. They can help people make sense of large bodies of information, generate new material, assist with decisions, coordinate workflows, personalise experiences, and operate across systems. AI systems are true socio-technical systems that participate and contribute.

But AI also introduces uncertainty. Responses are unpredictable. Sometimes, the system invents information or oversteps its role, acting on incomplete information while displaying an air of unnerving confidence. It exposes sensitive data, relies on the wrong source, or misses situations that require human judgement. A polished demonstration or prototype creates confidence under certain scenarios, yet struggles when faced with the complexity of real work.

This is why design matters.

Speed does not equal quality

Much of the current discourse around AI focuses on acceleration. The emphasis is on how quickly systems can be built, how easily interfaces can be generated, how existing workflows can be automated, and how traditional design and development processes can be compressed. There is a sense that the primary value of AI lies in speed: doing more, faster, with fewer people and fewer steps.

There is truth in this. AI can accelerate many parts of product development. It can help synthesise research, generate ideas and create working prototypes in code. Used well, it can significantly increase the speed and breadth of design exploration.

Yet, speed does not equal quality. Speed changes what teams are able to do, but it does not answer the more fundamental questions that shape whether a product will matter. What problem is being solved? What value is being generated? Whose life or work is being changed? What assumptions are being built into the system? What future is being made more likely? These are design questions, and they become all the more important, when the tools of production become more powerful.

Design, in this sense, is neither making something look appealing, nor a final layer applied after the important decisions have already been made. At its deepest level, design is the deliberate act of imagining, shaping, and testing possibilities before committing resources to action.  It is a way of understanding situations, framing problems, exploring alternatives, and moving from existing conditions toward preferred ones.

The activities of design are therefore forms of planning, but they are also forms of inquiry. Designers use sketches, diagrams, prototypes, scenarios, models, maps, specifications, and stories to explore ideas that do not yet exist. These are often know as design ‘artefacts’: tangible representations that allow emerging ideas to be shared, examined, questioned, and developed.

Such artefacts appear in every type of design discipline, from architecture and service design to engineering and software development. They are simplified representations of possible futures. They allow teams to examine complexity, compare alternatives, create alignment, and learn from failure before those failures become costly. Design occupies a unique position between thought and action: it makes the future thinkable, discussable and testable before it becomes real.

This understanding of design has deep roots. Herbert Simon described design as the act of devising courses of action aimed at changing existing situations into preferred ones. Horst Rittel showed that many important challenges are ‘wicked’ problems: complex situations whose causes, boundaries, and solutions are contested and constantly changing. Donald Schön described design as a reflective conversation with a situation, in which designers learn by making, observing, and revising their understanding as they go.

Taken together, these perspectives suggest that design is best understood as a practice of intentional change. Design helps people make sense of complexity, explore what could be, and act with judgement when certainty is impossible.

In the Age of AI, the scarce resource is judgement

The danger of purely experimental AI development, including what is sometimes called vibe coding, is not simply that it produces poor code. Often it produces decent code remarkably quickly. The deeper danger is that it can encourage teams to mistake the act of building for the act of understanding. A product’s success rarely depends on how quickly it was created. It depends on whether it addresses a meaningful need, fits into people’s lives and workflows, and creates outcomes that matter.

This is precisely why design remains essential in the age of AI. Generative models can accelerate the production of software, content, and services, but they cannot determine whether a problem is worth solving, whether a solution creates value for people, or whether its consequences are desirable. AI can help us build faster; design helps us build wisely.

As the cost of creating software falls, the scarce resource is no longer implementation alone. It is judgement. The organisations that succeed with AI will not necessarily be those that generate the most code, but those that best understand the problems they are trying to solve, the systems they are shaping, and the futures they are helping to create. In a world where almost anyone can build, design becomes the discipline that helps us decide what is worth building at all.

A system can be built quickly and still address the wrong problem. A polished interface does not guarantee a useful product. Strong model performance does not guarantee a good fit with real workflows. Fluent responses do not guarantee trust, sound governance, or operational value. Technical feasibility alone does not make a system appropriate for the people, organisation, or environment where it will be used.

AI’s real significance goes beyond building today’s products more quickly. It changes the kinds of systems we build. These systems respond, retrieve information, reason through problems, work with people and other systems, and take action in ways traditional software never could.

And new kinds of systems require new ways of designing.

How technology paradigms evolve

To understand why, it helps to consider how technologies evolve. Across different domains – from radio and television to personal computing, the internet, mobile devices, and now artificial intelligence – similar patterns can be observed. New technologies tend to move through phases. In the early stages of the paradigm, progress is driven primarily by engineering concerns and activities. The central question is what can be made possible. The people closest to the underlying technology explore its capabilities, push its limits, and demonstrate what it can do.

This early phase is often energetic, experimental, and chaotic – full of prototypes, speculative use cases, technical breakthroughs, and failed experiments. There are few established patterns, and fewer mature methods. Value is often tied to technical performance and first-mover advantage.

Over time, however, the limits of a purely technical focus become apparent. As the technology becomes more widely available, differentiation shifts away from capability alone and towards application. The question is no longer simply what the technology can do, but what it is for. 

• Who does it serve? 
• What problems does it solve? 
• What new meanings does it create? 
• How does it fit into people’s lives, work, organisations, and cultures? 

This is where design becomes central.

The most successful products and services rarely succeed because the technology is impressive. They succeed because people find them useful. They fit real needs, support organisational goals, respect real world constraints, and earn a place in everyday work. Design gives the technology a clear purpose and a reason to exist.

Eventually, the paradigm stabilises. Patterns become standardised. Practices become systematised. Organisations optimise for efficiency, scale, and predictability. This brings consistency and growth, but often narrows the design space. What was once exploratory becomes procedural. What was once novel becomes expected. Then, eventually, a new paradigm begins.

AI is currently in the early stage of this cycle. It is still largely engineering-led. The conversation is dominated by model performance, infrastructure, tooling, benchmarks, and technical capability. This work is essential. Without it, the paradigm would not exist.

Capability alone does not create value. As organisations move beyond experimentation, attention shifts to different problems. Teams need a clear purpose for the system. They need to understand who will use it, the context in which it will operate, the knowledge it should rely on, the decisions it should support, and the actions it should take. They need clear boundaries between human and AI responsibility, a good fit with existing workflows and operating models, and a plan for how the system will evolve over time. These are not merely technical questions, but design questions, and this is where many AI initiatives struggle.

How do you get from AI idea to production-ready system?

Organisations often start with a promising idea for an AI product. The goal might be to support customer service, improve knowledge search, assist research, summarise cases, review compliance, guide decisions, or help colleagues with everyday work. The concept usually sounds compelling, and an early prototype is often straightforward to build.

But between the idea and a production-ready system lies a large gap.

The team must define the system’s purpose, identify users and affected stakeholders, understand the workflows that create business value, decide what the AI knows and which sources it trusts, and determine how work is coordinated across models, tools, people, and systems. Clear expectations for behaviour are equally important. The AI needs rules for explanation, uncertainty, and escalation. Interactions should support trust, control, accessibility, and appropriate reliance. The wider ecosystem, legal obligations, organisational policies, safety requirements, and security constraints all shape the design. Monitoring, governance, and continuous improvement will complete the picture.

Without a design framework to guide, these decisions are often made implicitly. And when they remain implicit, they are shaped by unconscious assumptions, organisational biases, technical defaults, and local incentives. The system still has a purpose, but it may be poorly understood, or informed by false assumptions. It still creates meaning, but that meaning may be accidental. It still produces an experience, but that experience may be inconsistent, confusing, or harmful.

Purpose, meaning, and experience are always present in a system. The question is whether they are consciously designed.

Designers bring these questions into the open. They uncover assumptions, study real users in real contexts, turn abstract ideas into something people discuss and test, and give structure to uncertainty. Designers help teams focus on what matters and bridge the gap between human needs, organisational goals, and technical implementation.

This is why design matters in the age of AI.