The job that quietly disappeared
In 2021, IKEA introduced Billie, an AI assistant to handle routine customer service questions. Over the next two years, according to Ingka Group, the largest IKEA franchisee, Billie handled 47% of customers' queries to its call centres.
Ingka did not respond by cutting its call centre teams. It trained 8,500 call centre workers as interior design advisers, offering customers paid remote design consultations by phone and video. Remote channels, including that service, generated around €1.3 billion in the 2022 financial year.
The employees remained. The job they had been hired to do largely disappeared.
This is the emerging pattern of the AI economy: less mass unemployment, more the gradual erosion of traditional roles.
Jobs are bundles of tasks
Public debate about AI often focuses on a single dramatic question: will machines replace workers?
What is happening inside organisations is subtler. AI rarely eliminates whole occupations overnight. It targets the tasks that make them up. A job is a bundle of tasks: analysing data, drafting reports, answering customers, organising information.
AI starts to unbundle that bundle.
The International Labour Organization's 2025 analysis of generative AI exposure found that around one in four workers globally are in occupations with some exposure to generative AI, but only 3.3% of global employment falls into the highest-exposure category. Because most occupations consist of tasks that still need human input, the ILO concluded that transformation of jobs, rather than their replacement, is the most likely effect.
So the impact is felt less through immediate redundancies and more through the gradual reorganisation of work. An analyst may stop spending hours gathering data and building spreadsheets because AI now does those steps. A customer support agent may rarely see a routine enquiry because a conversational system resolves it first.
The job title remains. The work changes underneath it.
Over time, this hollows out the role. Tasks disappear piece by piece until the original definition of the job stops describing what the person actually does.
Machines take the structured tasks, people keep the context
The distinction matters. AI does not replace jobs all at once. It replaces the activities that make them up, unevenly.
Some tasks move quickly, while others remain stubbornly human. In financial services, investment analysts rely on AI to process large datasets, generate forecasts and surface patterns, but the analyst still interprets the results, judges their strategic meaning and advises clients. In healthcare, radiologists use AI that pre-reads scans and highlights possible anomalies, but the diagnosis remains a human responsibility.
Across industries, the pattern repeats: machines handle structured tasks and people handle context. Roles shift from production to oversight. A support agent becomes a resolver of complex cases. A marketer becomes a curator of AI-generated campaigns. A financial analyst becomes a strategist interpreting machine-generated insight.
The worker stays employed. The original job, the collection of tasks that once defined it, has quietly collapsed.
From job design to task architecture
For organisations, this challenges some of the most basic assumptions of workforce management.
Most companies still organise work around job descriptions that define responsibilities, pay, progression and reporting lines. When tasks change faster than roles, those structures become unreliable. Leaders may believe they are managing a stable workforce when the underlying shape of the work has already moved.
The answer is not just reskilling. It is redesigning how work is organised.
Forward-looking organisations are starting to think in terms of task architecture rather than job architecture. Instead of defining roles first and assigning tasks second, they analyse the tasks needed to deliver an outcome and then decide how those tasks should be split between people and machines.
In these models, AI does much of the structured execution, processing information, generating outputs and coordinating workflows, while people provide judgement, governance, creativity and interpretation. How that changes the organisation as a whole, from job descriptions to outcome-based teams, is explored in Beyond the Job Description.
A task-level workforce review
Workforce plans built on headcount by job title will miss this shift. A task-level review shows it. It can be run for one role family at a time.
- List the tasks. For each role in the family, list the ten to fifteen tasks that take most of the time, with a rough share of the working week for each.
- Classify each task. Mark it as one AI can largely do, one AI can speed up with a person still responsible, or one that stays human.
- Estimate the time released. Add up how much of the working week moves to AI over the next twelve to eighteen months, based on tools that exist rather than promises.
- Decide where the time goes. Choose deliberately: higher-value work in the same role, new roles (as IKEA did with remote design), or a smaller team over time.
- Build the paths. Identify the skills the new work needs and the training and internal moves that get people there.
- Revisit it. Repeat the review every six months, because the classification changes as the tools improve.
The value of the exercise is that it replaces a vague debate about whether AI will take jobs with a specific picture of which tasks are moving, and what the organisation will do with the time.
The end of the static career
For individuals, the shift can be disorienting. For generations, careers were built around roles: accountant, analyst, marketing manager, customer service agent. Professional identity was tied to the tasks those titles implied.
When machines take on a growing share of the tasks within a role, the profession changes. The analyst spends less time analysing. The writer spends less time writing. The service agent spends less time answering questions. The human role centres on supervision, interpretation and decision-making.
In practice, people may spend more time reviewing AI-generated work than producing it, and guiding automated systems rather than doing the tasks those systems perform.
That does not have to diminish human work. Often it elevates it. But it changes which skills matter. Judgement, creativity, ethical reasoning and contextual understanding become more valuable because machines struggle with them, while routine expertise becomes less distinctive.
Career progression may depend less on mastering a fixed set of tasks and more on learning to work effectively with intelligent systems. The professional of the future is not only a specialist. They are an orchestrator.
The disappearing job is the real story
The narrative about AI swings between optimism and alarm: either machines will replace everyone, or they will simply make work more productive. Both miss the deeper change.
AI is dissolving the structure of jobs themselves.
Employment will continue, but the definition of work inside organisations will change as tasks migrate from people to machines. The question for leaders is not only whether AI will reduce headcount. It is whether they understand how the work is changing.
Those who redesign their operating models around tasks, capabilities and collaboration between people and AI will capture the productivity gains. Those who keep managing work through outdated job structures may find themselves measuring roles that have ceased to exist in practice.
The future of work will be defined less by who holds a job than by how the work itself is redesigned.
