In the last volume of this series, we looked at how services became the infrastructure behind modern life, so that could lead us to understand how AI will impact our future. When starting this volume, I’d love to introduce a question to you:

What happens when AI starts operating inside that infrastructure?

To answer this, we first need to define what service design is, what it is for, and why it is growing as a career in 2026.

“A service is an organised system of actions, decisions, people, technology, rules, and resources that work together to produce an outcome for someone over time. Service design is the practice of shaping that system so the outcome can be delivered clearly, fairly, sustainably, and reliably.”

With AI entering this system, the role becomes more critical because AI may start influencing not only how the service is experienced, but how decisions inside the service are made.

And this is starting to show in the job market. The data is not perfect, because service design is still not measured as cleanly as product design, UX, or software roles. But the signal is there. In the UK, IT Jobs Watch recorded 159 permanent jobs citing Service Design in the first half of 2025 and 426 in the same period of 2026, which is a rise of around 168% in one year. The median salary for jobs citing Service Design also moved from £58,800 in 2024 to £65,000 in 2026. On LinkedIn, there are now 28,000+ Service Designer jobs worldwide, with more than 14,600 listed at mid-senior level. In the UK Government Digital and Data framework, service design is also defined across six role levels, from associate service designer to head of service design.

So, having that in consideration, what’s the difference? How do we differentiate Product Design and Service Design?

A product usually has a clearer boundary. It may be an app, a tool, a platform, a dashboard, a checkout, a feature and so on... The work is often about helping someone do something better, faster, more confidently, or with less confusion.

That work matters.

But a service is not contained in the same way.

A service is everything that needs to happen for an outcome to be delivered. It may include a product, but it also includes people, policies, rules, operations, identity checks, suppliers, caseworkers, data flows, payments, complaints, regulations, waiting lists, and exceptions. It can essentially happen outside your product scope.

Once AI leaves the product and enters the service, we are no longer only talking about productivity, revenue and improvement.

We are talking about the conditions under which decisions are made about people and the consequences for their lives.

This is the kind of thinking I go deeper into in ”System Thinking in Service Design Workbook”. If you are trying to become a Service Designer, I’d high recommend this workbook. It helps readers understand services beyond the visible experience, looking at the decisions, operations, policies, systems, teams, risks, and hidden structures that shape whether an outcome can actually be delivered in the real world.

Most of the media still talks about AI as a tool for productivity. It helps people write faster, summarise documents, generate ideas, analyse data, automate tasks or reduce costs. That story is not wrong, but it is incomplete. McKinsey’s 2025 global AI survey found that 88% of organisations now use AI in at least one business function, up from 78% in the previous year. But the more important question is not only whether organisations are using AI. It is where AI is being placed.

Once AI enters the operational layer of a service, it stops being only something a team uses. It starts becoming something the service depends on.

This is the moment where AI stops being only an assistant and becomes an active participant.

So the real challenge is not whether AI can make services faster.

It is whether we can still understand, challenge, and protect what happens to people once AI becomes part of the decision itself.

But before we go too far into AI, I think we need to be honest about something.

Services were already making decisions before AI arrived.

A welfare service already decides who is eligible for support. A healthcare service already decides who is urgent, who waits, and who gets referred. An immigration service already decides whose evidence is enough. A bank already decides which transaction looks suspicious. A tax system already decides which cases deserve attention. An insurance company already decides who is high risk. A hiring process already decides who moves forward and who disappears before interview.

So AI is not entering a neutral system.

AI is entering services that already contain rules, thresholds, categories, risk models, evidence requirements, operational pressures, and assumptions about people.

So the danger is not only that AI may make decisions. The danger is that it may take the decision logic that already exists inside a service and make it faster, more scalable, less visible, and harder to challenge. That is why the conversation cannot only be about accuracy or efficiency. It also needs to be about visibility, accountability, and whether people can challenge the system when it gets them wrong.

A useful book to read alongside this section is Weapons of Math Destruction by Cathy O’Neil. It helps explain how scoring systems, models, and automated classifications can appear neutral while quietly shaping access to jobs, loans, education, insurance, healthcare, and opportunity.

So, what is a human gate?

A human gate is not a person casually “checking” what the system says. It is a designed moment of responsibility, where someone has the authority, time, evidence, and obligation to review the decision before harm happens.

This matters because we already have warnings.

In Australia, the Robodebt scheme scandal in 2022 used automated income averaging to raise welfare debts, affecting hundreds of thousands of people and wrongly recovering large sums before the scheme was exposed as unlawful. The issue was not only that the calculation was wrong. It was that the system created debt at scale, while people had to carry the burden of proving the system wrong. (Source: Robodebt Gov.Au)

In the Netherlands, the childcare benefits scandal in 2018 showed something similar. An algorithmic risk system helped flag families for suspected fraud, and tens of thousands of mostly low-income parents and caregivers were falsely accused. Amnesty reported that racial profiling was built into the design of the system, with ethnic minority families disproportionately affected. (Source: Amnesty International)

Even in the UK, the 2020 A-level grading algorithm showed how dangerous automated judgement can become when it affects people’s futures. Almost 40% of students received grades lower than expected, leading to public outrage and the government eventually withdrawing the algorithmic grades. (Source: LSE)

These aren’t examples of “AI” in the modern generative AI sense. But they show the same potential service risk for the future.

When automated systems enter high-stakes decisions without meaningful human gates, the service can become faster at producing harm than people are able to challenge it.

This is why the EU AI Act matters. It entered into force in August 2024, with most rules applying from August 2026, and it treats AI through levels of risk. The point is important: some systems are not risky because they are technically impressive. They are risky because of where they are used, especially in areas such as employment, education, essential services, law enforcement, migration, and access to rights.

The more automated a service becomes, the more deliberately the human gate must be designed.

Otherwise, we are not building smarter services. We are building systems where responsibility disappears exactly when people need it most.

A resource designed around this kind of complexity is the ”System Thinking in Service Design“ Workbook. It helps readers look beyond individual service touchpoints and understand how service design shapes complex systems.

AI ethics is the attempt to define how artificial intelligence should be designed, used, monitored, and governed so it does not create harm, discrimination, manipulation, unsafe decisions, or loss of human accountability.

But the important point is that AI ethics only becomes real when it enters a context. A model is not ethical or unethical only because of how it performs technically. It becomes socially important because of where it is used, what it influences, who depends on it, and what happens when it is wrong.

The EU framework is important because it recognises that high-risk AI systems need requirements such as risk mitigation, high-quality data, transparency, clear information, and human oversight.

So, this is where AI ethics becomes service ethics too. Inside a service, bias is not just bias. It can become exclusion. Opacity is not just lack of transparency. It can become no route to challenge. Automation is not just efficiency. It can become reduced human discretion. Surveillance is not just data collection. It can become behavioural control. Error is not just technical failure. It can become delayed care, blocked money, rejected access, or unnecessary suspicion.

The real ethical question is not only how the model behaves.

It is what the service allows the model to change.

Because once AI starts affecting access, priority, evidence, risk, escalation, refusal, or support, ethics is no longer a principle on a slide.

It becomes the lived consequence of the service.

A useful book to read alongside this section is Automating Inequality by Virginia Eubanks. It is not a service design book, but it is deeply relevant to service ethics because it shows how data systems, policy algorithms, and predictive risk models can shape access to welfare, housing, and child protection. For this section, the value of the book is that it makes the ethical question concrete: automated systems do not only classify people.

If AI ethics becomes service ethics, then the future of work is not only about who can use AI faster.

It is about who can understand what AI changes once it enters real services.

This is where I think a lot of career advice becomes too shallow. “Learn AI” is not wrong, but it is incomplete. Most people will learn how to prompt, summarise, automate, generate, and speed up parts of their work. That will become normal but it will not be enough to make someone valuable in the future

The World Economic Forum’s 2025 report estimates that 22% of jobs will be structurally disrupted by 2030, with 170 million new roles created and 92 million displaced. But the most important signal is not only the number of jobs changing. It is the type of work becoming more valuable: analytical thinking, technological literacy, resilience, adaptability, leadership, systems thinking, and the ability to work with complexity.

And this is where service design has a stronger future than many traditional digital roles.

The valuable worker will not only be the person who knows how to use AI tools. It will be the person who understands where AI sits inside a service, what decision it influences, what data it depends on, who it affects, where it should stop, and how accountability survives when the system becomes more automated.

Some of these may become formal job titles. Others may become skills inside existing roles.

The future will reward people who can understand the consequences of making services more automated.

A practical resource for building this kind of thinking is the ”System Thinking in Service Design“ Workbook. It helps readers move past isolated touchpoints and understand how services operate as connected systems of decisions, people, processes, constraints, and outcomes.

So, where does this leave us?

I think the first step is to stop treating AI as something that only belongs to technical teams. Once AI enters a service, it becomes part of the service logic. It starts affecting how people are interpreted, prioritised, escalated, supported, questioned, or refused.

This means the future will need people who can sit between disciplines. But people who can understand enough of each layer to see what others may miss. Those are not necessarily designers.

The practical shift is this: instead of only asking how AI can improve the work, we need to ask what kind of service AI is helping to create. Because the danger is that organisations will use AI to make services look more efficient while making them less understandable, less accountable, and less human.

And this is where I think the opportunity is.

If you work in design, product, research, strategy, policy, operations, or transformation, your future value may not come from knowing every AI tool. It may come from understanding where AI should sit inside a service, where it should not sit, and what needs to be protected around it.

You do not need to become an AI engineer to have a role in this future. But you do need to understand the consequences of automation.

You need to understand how services make decisions, how people fall through gaps, how evidence is collected, how rules are applied, how exceptions are handled, and how accountability disappears when no one owns the whole system.

So the conclusion of this volume is simple:

Do not only learn how to use AI. Learn how to question what AI is being allowed to do and how to protect the people around it.

The future belongs to those who understand “People, AI, Systems, Strategy and Business”. It won’t be easy, but you can prepare for it.

If you are transitioning into Service Design, I have created this workbook to function as a comprehensive starting point. ”System Thinking in Service Design Workbook”

This workbook was designed to help professionals move beyond isolated outputs and understand how real services operate across systems, operations, governance, delivery ecosystems, and organisational complexity.

Inside the workbook, you’ll explore:

If you’d like to explore more topics around service, systems, AI, and organisational complexity, you can also follow me on Linkedin.

This volume was about understanding what happens when AI stops being only a tool and starts becoming part of how services interpret, route, prioritise, assess, and decide.

Once AI enters the operational layer of services, the question is no longer only whether the technology works. It is whether people can still understand what happened, challenge what went wrong, reach a human when needed, and trust the systems that increasingly shape access to care, money, work, education, identity, support, and rights.

Next, we move from AI as an active participant to the question of trust.

If services are becoming infrastructure, and AI is starting to make decisions inside that infrastructure, then trust becomes one of the most important design problems of the next decade.

The next volume will explore what makes a service trustworthy when decisions become automated, systems become harder to understand, and responsibility becomes distributed across organisations, platforms, policies, suppliers, models, and humans.