Will AI replace a job, change part of it or create different work around it? A convincing demonstration rarely answers all three questions.
AI could substantially change tasks by 2030, but a task's exposure to AI does not determine the future of the whole job. Employment also depends on whether organisations adopt the technology, how work is reorganised, what customers need and which activities someone is prepared to fund.
New work may emerge around AI-assisted processes. That does not guarantee new jobs, compensate every displaced worker or establish which occupations will grow.
For workforce planners, the useful starting point is therefore a specific role and its tasks. The future becomes easier to discuss when the steps between a technical possibility and an employment outcome remain visible.
What exposure indicators can tell us
The ILO's 17 April 2026 research brief on AI exposure indicators distinguishes technological susceptibility from labour-market outcomes. It explains that existing measures rely on task lists and judgments about capabilities, with limitations concerning adoption conditions, economic feasibility and changing demand.
An exposure score is therefore not a forecast that a particular employee will lose their job. Nor does it directly establish productivity gains or the exact training that person needs.
The World Economic Forum's Future of Jobs Report 2025 addresses a different question: employers' expectations through 2030. Published on 7 January 2025, it draws on a survey conducted in 2024. Its methodology describes 1,043 employer responses across 55 economies, focused on larger organisations.
The report considers several forces, including technology, economic conditions, demographics and the green transition. Its overall employment projections should not be relabelled as jobs created or destroyed by AI alone. Small enterprises and informal work are outside its principal scope.
These sources help frame the uncertainty. Neither supplies the future staffing plan for your organisation.
Start with the work inside a familiar role
Consider a fictional professional membership organisation that publishes briefings for its members.
A research coordinator finds relevant sources, checks dates, compares accounts, asks subject specialists for clarification and helps prepare a readable briefing. They also respond when a member says an important question was missed.
It is tempting to describe the job as “writing summaries.” That description leaves out much of the work.
Finding a document is different from deciding whether it answers the current question. Producing a paragraph is different from identifying a contradiction. Publishing a briefing also involves deciding what the membership needs to understand and what remains uncertain.
An approved AI assistant might support some of these activities under suitable conditions. Whether it does so usefully would need examination.
This is an original illustrative scenario, not a reported deployment or a prediction of a particular tool's capabilities in 2030.
Its purpose is to make a role visible as a combination of activities and responsibilities.
Which tasks could change?
The coordinator might use an assistant to organise suitable source notes, propose an initial comparison or suggest a draft structure.
Those possibilities are conditional. The information must be appropriate for the system, the proposed use must be supported and the output must meet the requirements of the work.
The change might reduce time spent producing an initial draft. It might also change the sequence: the coordinator could compare alternative structures earlier, then spend more attention resolving a disputed interpretation.
Alternatively, the assistant might add work without providing enough benefit. A plausible comparison could contain errors that take longer to identify than a direct reading of the sources.
The relevant question is what changes across the complete task, including preparation, checking and revision.
Some activities may remain recognisably similar. The coordinator still needs to understand a member's question and know when another person's expertise is required. Other activities may be combined or divided differently.
None of this settles whether the organisation needs more people, fewer people or different capabilities. It first establishes what the proposed change actually involves.
What additional work might emerge?
AI-assisted production can raise questions that someone must handle. In the fictional organisation, several possibilities deserve investigation.
Keeping the source collection usable
If a briefing process draws on an approved collection of material, someone needs to decide what belongs there, notice when it becomes outdated and correct the information the process relies on.
That work may already exist in another form. AI could make it more visible, change its frequency or require a different connection between research and publishing.
Calling it a new job would be premature. It might remain part of the coordinator's role or be shared with existing editorial and information-management colleagues.
Examining difficult outputs
A draft might combine two sources that use different definitions. Someone needs enough subject understanding to notice the mismatch and decide what further investigation is appropriate.
An established review process may already provide that capability. If the proposed system creates a different pattern of errors or a larger volume of material, the review arrangement may need to change.
That is a possible work requirement to test, not evidence of a guaranteed expansion in quality-assurance employment.
Translating members' questions into a workable service
Members may begin asking for different forms of help once a new service becomes available. Someone must establish whether those requests are useful, feasible and within scope.
The organisation might choose to offer a new kind of briefing. It might decide that a more personalised service would require expertise and resources it cannot provide.
The potential new work lies in understanding the need and designing an appropriate response. Demand for that work and the willingness to pay for it remain open questions.
These examples are possible changes around a fictional process. They are not a forecast of three new occupations.
A task is not automatically a paid role
An organisation may discover additional work without creating a position.
The activity could be brief, occasional, distributed among existing staff or included in a supplier's service. It could also be important enough to justify a dedicated role.
To distinguish those possibilities, examine the volume, continuity, expertise and authority involved. Who needs the work done? How often? What happens if it is not done? Which other responsibilities would compete for the same person's attention?
For the membership organisation, maintaining a small source collection may fit within existing editorial work. A substantially different service could require additional capability. Neither outcome follows simply from the presence of AI.
New titles can make a change sound more definite than it is. “AI workflow specialist” may describe a real bundle of responsibilities, a temporary project assignment or an aspiration with no funded work behind it.
Look for the responsibilities and resources before treating the title as evidence of an emerging career.
Follow the employment question beyond the first saving
Suppose the fictional organisation finds that assisted drafting reduces part of the briefing workload under the tested conditions.
Several choices remain.
It might publish more briefings. It might improve existing coverage, give specialists more time to investigate difficult questions or reduce an overloaded schedule. It might also seek to reduce staffing costs.
The technology does not select between those choices. Available resources, demand, service priorities and organisational decisions matter.
Nor would a reduction in one organisation establish the economy-wide effect. Other organisations may face different costs, needs and opportunities. Conversely, the existence of growing AI-related work elsewhere would not remove the consequences for a person whose role changes.
A responsible discussion should hold both points at once: aggregate possibilities are uncertain, and local transitions can still be consequential.
This article does not offer a headcount formula or a guarantee of a favourable outcome.
Expect differences within the same occupation
Two people with the same job title can perform different combinations of tasks.
One research coordinator may produce largely standardised updates from a stable set of materials. Another may spend much of the week investigating unfamiliar questions with specialists. Their opportunities and constraints will differ.
Organisations also differ in access, information quality, resources and the conditions under which a system may be used. A technically plausible activity may not be suitable or worthwhile in every setting.
Country and sector matter too. A global occupational category cannot describe every local service, labour market or institution. The role of informal work and smaller organisations deserves particular attention when applying findings drawn mainly from larger employers.
These differences argue for examining actual work before announcing a universal future for an occupation.
They do not imply that change will be minor. They explain why its form and consequences need to be investigated.
Make 2030 a planning horizon
A distant date can encourage certainty that the evidence does not support. It can also be useful if it helps leaders examine decisions they can influence now.
For the fictional membership organisation, begin by describing the present briefing process with the people doing it. Include the parts that are rarely mentioned in the job description: clarification, correction, coordination and handling unusual requests.
Then choose a bounded proposed change. Specify what the assistant would support, what people would still need to do and what new work might follow.
Test that proposal under appropriate conditions. Examine the complete workload and the usefulness of the finished service. Record which assumptions held and which did not.
The next decision could concern a revised process, a different learning need, an additional responsibility or stopping an unsuitable use. It need not begin with a prediction about the entire occupation.
This is a proposed planning approach, not a validated forecasting method. It makes the reasoning easier to examine while leaving uncertain outcomes uncertain.
Prepare people for a transition you can describe
“Learn AI” is too broad to serve as the complete development plan for a changing role.
For the coordinator, a useful learning objective might be comparing an assisted synthesis with its source material and explaining an important discrepancy. Another might concern maintaining the information the process uses.
A different role could require different preparation.
The human learning advantage offers a broader perspective: people need ways to acquire understanding, judge unfamiliar situations and adapt as the work changes.
That does not make learning a guarantee of job security. Organisations should distinguish a development opportunity from an actual employment commitment. They should also explain what has been decided and what remains under consideration.
Employee knowledge belongs in the planning process. The person doing the work may identify an overlooked task or a dependency that changes the proposed redesign.
Ask a future-of-work event to clarify the assumptions
A useful conference session can move beyond naming jobs that will supposedly disappear.
Give participants a fictional role such as the research coordinator. Ask them to identify one task that could change, one activity that may become more important and the assumptions between that change and a staffing decision.
Different answers can reveal different views of demand, quality, resources and the purpose of the service.
The exercise is an editorial proposal, not an evaluated intervention. A keynote can introduce the distinctions; a working session can connect them to a particular organisation.
Our guide to designing a corporate AI learning summit explains how to support that conversation with subsequent workplace learning.
The question behind a 2030 headline
Sources were checked on 28 October 2026. This article uses the ILO's public research-brief summary and the WEF's report overview and methodology. It is not a systematic review or a new employment forecast.
The fictional role and possible changes illustrate questions to investigate. They do not establish that the activities will expand or become separate paid occupations.
By 2030, AI may have changed the contents of many jobs. Understanding a particular transition requires more than knowing what a system can generate.
Before accepting a prediction, ask: Which tasks change, what work follows, and what still has to happen before that becomes an employment outcome?
Bring that discussion to your next future-of-work or leadership event. Explore Learning Coach's international keynotes and request a tailored proposal.
References
Merola, R., Ernst, E., Samaan, D., del Rio-Chanona, M., and Teutloff, O. Workers' exposure to AI: What indicators tell us—and what they don't. ILO, 17 April 2026. Public brief summary and measurement limitations consulted.
World Economic Forum. The Future of Jobs Report 2025, 7 January 2025, and report methodology. Employer expectations across multiple trends; survey conducted in 2024.
