AI Is a Work Design Problem

Most organizations still approach new technology in roughly the same way: choose a tool, configure it, introduce it to employees, train people to use it, and manage the resulting change. The underlying assumption is easy to miss. The technology is treated as the thing being designed, while the organization and the people working inside it are expected to adapt.

Artificial intelligence makes the limitations of this approach much harder to ignore.

AI doesn’t simply give employees a new tool. It can change which tasks people perform, what information they have access to, which decisions they are allowed to make, how their performance is measured, and where responsibility sits when something goes wrong. It can remove tedious work, but it can also remove opportunities to develop expertise. It can give people greater autonomy or subject them to greater surveillance. It can make an existing job easier, fundamentally change that job, or eliminate it.

In other words, implementing AI is also designing work.

That distinction matters because much of the current conversation about AI and employment treats technology as though it independently determines what happens next. We ask which jobs AI will replace, which tasks it will automate, and which skills workers will need to remain valuable. The worker is positioned downstream from the technology: AI changes, and people must figure out how to keep up.

But technology creates possibilities. Organizations decide which possibilities become reality.

Imagine an AI system can perform 30 percent of the tasks currently assigned to a particular role. The most obvious response might be to calculate how many fewer people the organization now needs. But that is only one possible design decision. The organization could instead use the additional capacity to serve more customers, improve quality, reduce working hours, expand the responsibilities employees find meaningful, spend more time solving difficult problems, or create entirely different roles.

The technology hasn’t dictated any of these outcomes. It has changed the conditions under which work can be designed.

We have been here before

The idea that people should not simply be reorganized around technology isn’t new. It sits at the center of sociotechnical systems theory, which emerged from research into the relationship between people, technology, and work beginning in the mid-twentieth century.

One of the people who carried that thinking into information systems design was Enid Mumford, who rejected the assumption that experts should design technological systems and then expect employees to accommodate them. Her work treated the technical system and the social system—the people, roles, relationships, and organization of work—as interdependent.

The principle that emerged from this tradition is known as joint optimization. Instead of optimizing technology and then dealing with its human consequences, organizations should design the social and technical parts of work together so that the whole system performs better.

That sounds remarkably contemporary when applied to AI.

The conventional question is: How can we use AI to make this process more efficient?

A sociotechnical question is broader: How should people, work, and technology be organized together now that this capability exists?

Those questions can produce very different futures.

Consider a customer-service team. An AI system might summarize calls, retrieve information, draft responses, and handle simple requests automatically. If the organization treats the project primarily as automation, the objective may become reducing the number of people required to process the same volume of work.

But another organization could ask what humans are particularly good at and redesign the work around that. AI could handle routine transactions while employees spend more time resolving unusual cases, exercising judgment and building relationships with customers. Information previously available only to specialists might become accessible to frontline employees, allowing them to resolve problems without escalating them through layers of management. The same technology could support a job with greater autonomy and skill—or one characterized by greater monitoring and work intensity.

Neither outcome is inherent in the AI.

It is designed into the work system surrounding it.

Research by Sharon Parker and Gudela Grote on automation, algorithms and work design reinforces this point. Digital technologies can increase or decrease autonomy, skill use, feedback, social connection, well-being and performance depending on how organizations redesign work around them. Their research challenges the assumption that we can understand the consequences of automation simply by examining the technology's capabilities.

This is why “Will AI replace this job?” is often the wrong design question. Jobs are bundles of tasks, responsibilities, relationships, and decisions. When technology changes some of those components, organizations can reconsider how the remaining pieces fit together.

The more useful question is: What should this job become?

The designer’s role has to change too.

This brings us back to the larger argument of this series.

If human-centered design is expanding beyond bounded products, if wicked problems require us to examine how problems are framed, and if systems contain people with different knowledge, interests, and power, then designers working with AI cannot limit themselves to making AI products easier to use.

The interface matters. But so does everything around it.

A designer working on an AI system might need to understand how work currently happens before deciding what to automate. That means observing not only formal processes but also the workarounds, judgment calls, and informal coordination that keep an organization functioning. They may need to understand why particular decisions sit with managers, what knowledge employees develop through experience, where accountability lives, and what happens when the technology is wrong.

The resulting design work may extend well beyond the screen. A prototype could be a different division of responsibility between a person and an AI system. It could be a new decision-making process, a redesigned role, an escalation path, a governance rule, or a different way of measuring performance.

This doesn’t mean product designers suddenly become experts in organizational design, labor economics, cybersecurity, ethics, and corporate governance. Complex systems require multidisciplinary work because no designer can understand them alone.

But designers do bring something valuable to that work. We know how to investigate lived experience, surface knowledge that formal processes overlook, make complex relationships visible, bring different perspectives into the same conversation, and make possible futures tangible enough to evaluate before they become permanent.

Those capabilities become more important, not less, as technology becomes more powerful.

The future of design is not another framework.

This is why I am skeptical when the evolution of design is framed as the death of human-centered design or the Double Diamond.

The challenge facing designers isn’t finding the diagram that finally replaces the old one. It is becoming more sophisticated about what we’re designing, where we draw the boundary, and what kind of problem we’re actually facing.

Sometimes the appropriate intervention is a better interface. Sometimes it is a redesigned service. Sometimes it is a workflow. Sometimes it is the relationship between several departments. And increasingly, as AI enters organizations, what needs to be designed is the relationship between people, work, and technology itself.

Human-centered design remains essential because we still need to understand how people experience those systems. But understanding people cannot end with asking whether they can use the technology successfully. We also have to ask what the technology allows them to do, what it prevents them from doing, how it changes their authority and autonomy, what new risks it creates, and who receives the value it produces.

AI gives organizations an extraordinary opportunity to rethink work. It could remove tasks people never wanted to do, make expertise more accessible, enable smaller organizations to accomplish things previously available only to large companies, and give people greater capacity to focus on work requiring judgment, creativity, and relationships.

It could also intensify work, expand surveillance, concentrate decision-making and use every productivity gain primarily to reduce labor costs.

Both futures are technologically possible.

Which one we create is a design decision.

The future of design therefore isn’t about choosing between human-centered design and systems design. It is about becoming capable of working across both: understanding individual human experience while also understanding the organizational and technological systems producing it.

For the last decade, the technology industry largely asked designers to become better at designing products.

The next decade will require us to become better at designing how people and technology work together.