Key takeaway
- Human-centered design explained: four principles, the ISO 9241-210 activities, methods by situation and how to apply them to enterprise AI agents.
Human-centered design is a way of building products, services and systems around the people who use them and the context they work in, then checking each version with those people before it ships. For an enterprise it is a risk control as much as a design method: it finds the wrong problem early, before a team funds the wrong solution.
The standard reference is ISO 9241-210, which sets out principles and activities for human-centred design of interactive systems. If you also want the broader creative method, read our guide to the design thinking process; this guide stays on the human-centred principles, the ISO activities and how to apply them to software and AI agents.
What is human-centered design?
Human-centered design (HCD) puts people and their context first. Interaction Design Foundation, citing the work of Don Norman, describes four key aspects: focus on people and their context, solve the right (root) problems, treat everything as a system, and make small and simple interventions that are tested and refined (source: Interaction Design Foundation, read 13 Nov 2026).
It goes further than "user-friendly". The people affected by a design include the end user, but also the support agent, the administrator, the reviewer and anyone who lives with the outcome of a decision the system makes.
What are the four principles of human-centered design?
1. Focus on people and their context
Name the real people, the tasks they are trying to finish, and the place, device and time pressure they work under. "The user" is not specific enough. A claims handler on a shared screen and a field engineer on a phone need different designs, even when they use the same system.
2. Solve the right problem, not the symptom
Observe the work and talk to the people doing it before you write requirements. Treating a symptom lets the problem return, which is why research comes before solutions. NN/g advises spending most of a limited research budget early, when findings can shape the product most (source: NN/g UX research cheat sheet, read 13 Nov 2026).
3. Treat everything as a system
A polished screen does not rescue a broken journey. If a payment fails and money is deducted but the message to the customer is unclear, the whole experience fails, however good the interface looked. Design the journey end to end, including errors, handovers and support.
4. Make small, tested changes
Prototype the cheapest thing that tests your thinking, such as a sketch or paper prototype, put it in front of real people, then refine. NN/g suggests starting with paper prototypes, moving to interactive ones, and watching whether people complete tasks without errors rather than asking for their opinions.
What does ISO 9241-210 say about human-centred design?
ISO 9241-210:2019, "Ergonomics of human-system interaction, Part 210: Human-centred design for interactive systems", gives requirements and recommendations for human-centred design principles and activities across the life cycle of computer-based interactive systems. It replaced the 2010 edition. A standards listing on genorma shows it as confirmed in May 2025; confirm the current status on iso.org before you cite it in a policy.
As summarised by Userfocus, the standard lists six principles:
- The design is based on an explicit understanding of users, tasks and environments.
- Users are involved throughout design and development.
- The design is driven and refined by user-centred evaluation.
- The process is iterative.
- The design addresses the whole user experience.
- The design team includes multidisciplinary skills and perspectives.
Source: Userfocus on ISO 9241-210, read 13 Nov 2026. The principles are paraphrased from that article, not from the standard text, which is paid.
What are the phases of human-centered design?
Sources differ on names and counts. The ISO route is plan the process, then understand and specify the context of use, specify the user requirements, produce design solutions, and evaluate the design, repeating the loop until requirements are met (steps as listed in a Rheinwerk Computing guide to ISO 9241-210, read 13 Nov 2026). Many teams use friendlier labels. The table maps them.
| ISO 9241-210 activity | Common team label | What the team produces | Typical methods |
|---|---|---|---|
| Understand and specify the context of use | Empathise and define | Who the users are, their tasks and environment | Interviews, field observation, user research methods, empathy mapping |
| Specify the user requirements | Define and reframe | A clear problem statement and measurable needs | Affinity mapping, journey maps, user stories |
| Produce design solutions | Ideate and prototype | Sketches, paper and interactive prototypes | Brainstorming, low-fidelity prototypes |
| Evaluate the design | Test | Evidence of what works and what does not | Usability testing, accessibility review, benchmark tasks |
Release is not the end of the loop. NN/g includes a "listen" stage for monitoring after launch, using surveys, analytics, search logs and support feedback. Use the ISO loop to decide what to build, and the listening stage to learn how it behaves in use.
Teams that run the loop for the first time often ask which activity to do first if time is short. NN/g's answer for an existing system is qualitative, think-aloud usability testing, because analytics show what is happening but usually not why.
How do you apply human-centered design in an enterprise?
Start with the decisions, not the screens. Pick one journey where failure is costly, name the people on both sides of it, and agree what a good outcome looks like before design starts. Then choose the lightest method that answers the open question.
| Situation | Best first move | Why it fits |
|---|---|---|
| New product, unclear need | Field study and stakeholder interviews | Checks the problem is worth solving before build (NN/g "discover") |
| Concept chosen, design open | Paper prototype and task testing | Cheap, fast and finds usability problems early |
| Live system with complaints | Think-aloud usability test plus support-log review | Finds why users struggle, not only where |
| Redesign needs a baseline | Benchmark tasks (time, completion, errors) | Lets you compare before and after |
| Large estate, many teams | Shared principles and a design review gate | Keeps decisions consistent across teams |
Visual and interaction choices follow the research, not the other way round. Our guides to user experience design and visual design principles cover the craft once you know what people need.
How does human-centered design apply to AI agents?
When software acts on a person's behalf, the human-centred questions get sharper. Who can see what the agent did? Who can stop it? What does the person see when it is unsure or wrong? The system view matters most here: the interface is only one part of a journey that includes approvals, exceptions and audit.
Our guide to human-centred AI and its design principles covers this side in more depth. In practice, design the oversight points with the same care as the main task: a clear handover to a person, an explanation the reviewer can act on, and a record of what happened. ElixirClaw is an agentic OS platform to build, orchestrate and govern AI agents and automation workflows while keeping humans in control, and AgentGRC keeps autonomous agent actions inside policy, with the audit trail risk teams need, according to their own sites.
What are the trade-offs, and when should you not use this approach?
| Trade-off | What it means in practice |
|---|---|
| Time and budget | Research takes time, but skipping it risks solving the wrong problem. Scale the method to the risk of the decision. |
| Small samples | Qualitative studies show why people struggle, not how often. Pair them with analytics for scale. |
| Stakeholder pressure | A senior opinion is not evidence. Use test results to settle disputes. |
| Fixed-scope contracts | Iteration needs room to change. Agree the evaluation criteria up front. |
| Purely technical changes | A back-end migration with no user-visible effect needs less research, but check support and operations staff are not affected. |
How do you adopt human-centered design step by step?
- Choose one high-value journey and name its owners.
- Observe and interview the people in it, including support and operations staff.
- Write the problem and the success measures in plain words.
- Prototype the cheapest version and test it with real users from different groups.
- Fix what testing shows, test again, and only then build.
- After release, listen: review support logs, analytics and feedback on a schedule.
Pair this approach with design thinking when the problem is open, and with user research methods when you need to choose between interviews, field studies and tests.
To map one decision your agents should make, and the context and policy it depends on, book a Context Workshop. See what happens in the workshop.
Frequently asked questions
What is human-centered design in simple terms?
It is designing products, services and systems around the real needs and context of the people who use them, and testing each version with them. Interaction Design Foundation lists four key aspects: focus on people, solve root problems, treat everything as a system, and make small, tested interventions.
What are the four principles of human-centered design?
Focus on people and their context, understand and solve the right (root) problems, treat everything as a system, and make small and simple interventions that you test and refine.
What is ISO 9241-210?
ISO 9241-210:2019 gives requirements and recommendations for human-centred design principles and activities across the life cycle of interactive systems. It replaced the 2010 edition. Confirm its current status on iso.org.
What is the difference between human-centered design and design thinking?
Human-centered design is the principle of designing around people and their context, with an ISO standard behind it. Design thinking is a creative method that many teams use to put those principles into practice. The two overlap, so most teams use both.
Which activity should we do first if time is short?
NN/g recommends qualitative think-aloud usability testing for an existing system, because analytics show what is happening but usually not why.
Explore the XenonStack platform family: ElixirData · ElixirClaw · ElixirHub · NexaStack · MetaSecure · Neural AI · AgentGRC · AgentSRE · AgentAnalyst · AgentInstruct · AgentRAI · AgentScrum · AgentSearch · AgentForce · AI Aviator · SAIF Aviator · CohortIQ · XenonStack AI · XenonStack UAE
