Generative AI has changed the conversation around digital products remarkably quickly.

Organisations are asking where AI can automate work, improve services, reduce costs and create entirely new experiences. Product teams are experimenting with capabilities that would have seemed improbable only a few years ago.

But amid this acceleration, there is a risk of beginning with the technology rather than the people who will encounter it.

The question becomes: “What can we do with AI?”

A human-centred approach begins somewhere else: “What are people trying to achieve, in what context, and where could technology genuinely help?”

That distinction is important.

What ISO 9241-210 gives us

ISO 9241-210:2019, Ergonomics of human-system interaction — Part 210: Human-centred design for interactive systems, sets out requirements and recommendations for human-centred design principles and activities across the life cycle of interactive systems. ISO reviewed and confirmed the 2019 edition in 2025, so it remains the current edition.

Its principles remain strikingly relevant. Human-centred design starts with an explicit understanding of users, tasks and environments; involves users throughout design and development; uses human-centred evaluation to drive and refine design; works iteratively; considers the whole user experience; and benefits from multidisciplinary perspectives.

These principles were not written specifically for generative AI. Yet they provide an unusually useful lens through which to examine it.

From user-centred interfaces to human-centred intelligence

Traditional digital product design often asks whether someone can successfully use an interface. AI introduces additional questions.

Can someone understand what the system is doing? Do they know when its output might be unreliable? Can they challenge, correct or override it? What happens when confidence in an AI-generated answer is misplaced? Where should automation stop and human judgement begin?

And perhaps most importantly: Should AI be involved in this particular problem at all?

This extends human-centred design beyond usability. It becomes a question of agency, understanding, accountability and trust.

Human-in-the-loop should be a design decision

“Human-in-the-loop” is frequently discussed as an AI governance mechanism. But it is also a design problem.

Simply placing a person somewhere in an automated process does not necessarily create meaningful human oversight. We need to understand what that person is expected to do.

Are they making the decision? Reviewing an AI recommendation? Checking an exception? Correcting an output? Taking responsibility for something they cannot realistically verify?

This is consistent with wider AI risk-management guidance. The NIST AI Risk Management Framework calls for human roles and responsibilities in human-AI configurations to be clearly defined and for human oversight processes to be documented. UK ICO guidance also warns that human reviewers can overestimate AI outputs through automation bias, and recommends monitoring hybrid human-AI systems and reviewer overrides.

Good human-centred design therefore examines these interactions deliberately. The goal shouldn't necessarily be maximum automation. It should be an appropriate relationship between human capability and machine capability.

Context becomes even more important

One of the most valuable ideas within ISO 9241-210 is understanding the context of use.

People don't experience technology in isolation. They use systems within organisations, environments, workflows, relationships and constraints.

An AI tool that performs impressively in a demonstration may behave very differently when introduced into an actual workplace. People may lack the time to verify its recommendations. Organisational incentives may encourage over-reliance. The available data may poorly represent particular users. Existing workflows may conflict with the new system. People may develop unexpected workarounds.

Understanding these realities requires research with the people affected — before, during and after implementation.

Evaluation cannot end at launch

Human-centred design is iterative. That principle becomes particularly important for AI-enabled products because their consequences may only become visible through use.

Evaluation therefore needs to extend beyond “Can people use it?” towards questions such as: Does it help them achieve what they need? Do people understand its limitations? When does it fail? Who is affected when it fails? Are people becoming appropriately reliant — or overly reliant — on it? Is the system creating the organisational outcome we intended?

NIST similarly treats AI risk management as a continuous lifecycle activity, including testing before deployment and monitoring systems in operation.

A technically successful AI implementation can still be a poor human system.

Start with people, not AI

Perhaps the most useful lesson from human-centred design is also the simplest. Don't begin with the solution.

Begin by understanding people, their goals, their environment and the problem. Then identify the opportunity. Then determine what should be designed. And only then ask what role AI — if any — should play.

For organisations exploring AI today, this changes the sequence from Technology → application → user to something closer to People → context → problem → opportunity → technology → experience → evaluation → impact.

AI may be an important part of the answer. Sometimes it may not be the answer at all. Recognising the difference is itself part of responsible innovation.

References

International Organization for Standardization. ISO 9241-210:2019, Ergonomics of human-system interaction — Part 210: Human-centred design for interactive systems. Current edition confirmed in 2025.

National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework (AI RMF 1.0), 2023.

UK Information Commissioner's Office. Guidance on AI and data protection — fairness in the AI lifecycle.

FOUR HUMAN / INSIGHT

At FOUR HUMAN, we believe emerging technology should begin with an understanding of people. Our work connects human-centred research, product and UX strategy with responsible approaches to AI and organisational change.

Intelligence. Experience. Identity. Impact.
FOR HUMAN.