The revealing question

Does your organisation design its AI solutions, or does it just engineer them?

Most organisations would say they design them. In practice, what they usually mean is that they make a series of technical choices: which model to use, which data pipelines to build, and so forth.

These decisions are necessary and they matter. But they are not Design.

Design, in the sense of whether an AI solution succeeds or fails in the real world, is something else entirely.

Design with a capital ‘D’ covers the intangible qualities that make an AI solution trusted, safe and customer centred. It addresses questions that cannot be answered by engineering alone.

Questions like:

• how people experience the solution,
• how AI changes behaviour and the business relationship, and
• how the solution aligns with brand values.

Design is the missing element in many AI initiatives, and it explains why so many technically-competent AI projects struggle once they leave the lab.

The limits of engineering-led AI

Engineering processes are exceptionally good at solving well-defined problems. They excel when requirements are stable, constraints are explicit, and success can be measured through clear performance metrics. Much of modern AI development has inherited this mindset.

But AI solutions do not operate in closed, predictable environments. They act within human, organisational, and societal contexts that are dynamic, ambiguous, and often contradictory. Customer expectations shift. Social norms evolve. Regulatory interpretations change. Organisational incentives conflict. Data drifts.

These considerations are not outliers. Rather, they are the real world.

When AI initiatives are driven primarily by engineering logic, these considerations are often treated as secondary concerns or “risks to be mitigated later.” The result is predictable. Systems that technically work nonetheless generate customer frustration, internal resistance, reputational damage, or regulatory exposure. Because AI systems operate at scale, these failures tend to be amplified rather than contained.

Broader enterprise patterns support this analysis. A 2025 S&P Global survey found that 42% of businesses scrapped most of their AI initiatives in the previous year, many before they even reached production. This reflects not a lack of technical capability but a failure to connect those capabilities with real operational and human requirements. Put simply, the projects were engineered competently but insufficiently Designed.

When newsworthy problems occur such as a chatbot offering bad advice and becoming a court case, or a chatbot refusing to answer questions that mention the brand’s name, the answer is not just to “fix this in the next release”. The root of the issue happened further back in the process, where Design was needed.

What Design actually covers

Design, properly understood, is not decoration and not usability polish applied at the end. It is a way of thinking and working that addresses complexity, uncertainty, and human consequence from the outset.

Design encompasses:

• Customer experience: not only in interface terms, but in how AI decisions are perceived, explained, contested, and trusted over time.
• Employee reception: including how AI reshapes roles, authority, accountability, and professional identity inside organisations.
• Societal and ethical impact: moving beyond compliance checklists, addressing fairness, power asymmetries, and long-term consequences.
• Regulatory and governance realities: interpreted not as static rules but as evolving expectations shaped by public scrutiny.
• Brand alignment: ensuring that AI behaviour reinforces, rather than undermines, what an organisation stands for.

These considerations are interdependent and often in tension. Optimising for one in isolation usually degrades another. Design is the discipline that works within these tensions rather than pretending they do not exist.

From engineered AI to Human-Centred AI

When Design is treated as a first-class concern, the result is Human-Centred AI (HCAI).

Human-Centred AI is not a slogan or an ethical add-on. It is a commercial outcome. Solutions that are trustworthy, responsible, reliable, and aligned with brand and values perform better in the market because people are willing to adopt them, rely on them, and recommend them. They fail more gracefully. They attract less regulatory friction. They generate fewer costly surprises.

Crucially, HCAI cannot be produced by engineering alone. It cannot be guaranteed by model accuracy, robustness testing, or governance frameworks in isolation. These are necessary ingredients, but they do not determine how an AI solution is experienced or judged in context.

Human-Centred AI must be Designed.

This is why organisations that focus solely on replicating “best practices” or following formulaic AI playbooks often struggle to differentiate. They are optimising execution while neglecting intent. They are building systems that resemble everyone else’s, and they encounter the same problems everyone else does.

Organisations that overlook HCAI increase the risk of poor ROI, high-impact customer rejection, and negative news coverage.

Why HCAI is hard

If HCAI were easy, it would already be ubiquitous. The reality is that even highly capable organisations find it difficult to implement AI in ways that consistently deliver value without unintended consequences.

Design is demanding work. It requires comfort with ambiguity and the ability to make informed decisions before all variables are known. It requires engaging with stakeholders who disagree, surfacing uncomfortable trade-offs, and resisting the false certainty that technical metrics can provide.

It also requires a combination of capabilities that are rarely found in one place:

• Design expertise, including the ability to frame problems, explore alternatives, and reason about human behaviour.
• Designerly thinking, which values iteration, learning, and sense-making over premature optimisation.
• AI literacy, sufficient to understand what AI can and cannot do, and how commercial decisions interact with technical constraints.
• Interdisciplinary collaboration, spanning engineering, strategy, policy, user experience, and organisational change.

This leads to a clear contrast:

Without Design: experience decisions are made late, risks are framed as exceptions, and success is measured by deployment.

With Design: experience decisions are made early (where positive impact is highest), trade-offs are surfaced and managed deliberately, and success is measured by adoption, trust and commercial value.

Research Through Design, and innovation

One of the most effective ways to practice HCAI is through Research Through Design. Rather than treating research as something that precedes implementation, this approach uses design activity itself to generate insight and knowledge.

By prototyping scenarios, exploring alternative futures, and testing assumptions through designed artefacts, teams can better understand what problem they are actually solving. This shifts the focus from “How do we build this system?” to “What should exist at all, and why?”

This is where innovation emerges—not as novelty for its own sake, but as a considered response to real human and organisational needs. Innovation, in this sense, is the disciplined act of designing better futures with AI, rather than extrapolating the present with more powerful tools.

Next steps

If you are currently implementing AI, it is worth pausing to ask a few uncomfortable questions. Who is this solution really for? What behaviours does it encourage or constrain? Where are decisions about trust, accountability, and impact being made implicitly rather than deliberately?

When these questions lack clear ownership, Design is often the missing element.

We work with organisations to surface these issues early and turn them into sources of clarity and value, not risk. As one example of HCAI in action, we have worked with Envolve Tech to bring intentional Design into their business processes, and you can read about the process and impact in the case study.

If you would value a conversation about how HCAI Design could shape your next phase of AI innovation, reach our team at [email protected] .