Artificial intelligence may help economies cope with ageing workforces, but the industries most exposed to demographic pressure are not necessarily those easiest to automate. That mismatch is the central finding of a BIS bulletin published on September 24, offering a more selective view of the productivity story behind the AI investment boom.
The study by Iñaki Aldasoro, Sebastian Doerr and Daniel Rees compares ageing and automation exposure across more than 130 economies. It finds that sectors with younger workers, such as finance, can be more amenable to automation than older, high-employment sectors such as agriculture and health. The result complicates the assumption that better software will neatly replace every worker an economy loses.
An economy is not one interchangeable workplace
A national workforce can shrink while the demand for particular services rises. Older populations may need more care, for example, even as fewer workers are available to provide it. Software that improves document processing does not automatically perform the physical and interpersonal tasks involved in delivering that care.
This makes the distribution of productivity gains important. A large improvement in one sector can help the economy, but it does not necessarily remove a bottleneck elsewhere. Workers, equipment and skills cannot always move instantly to the activity where demand is greatest. The speed of that adjustment affects how much of a technological gain becomes usable capacity.
For investors, the lesson is to examine the actual task being automated. A company selling a tool that saves administrative time is addressing a different market from one trying to replace physical labor. Both may have value, but their costs, adoption hurdles and customer needs differ substantially.
Exposure is not the same as deployment
An occupation can contain tasks that software could perform without a business being ready to reorganize around it. Implementation requires systems, training and a way to judge the quality of the output. In sensitive activities, errors can create costs that offset some of the time saved.
The same distinction applies to financial services. A model may help staff review information, but a regulated institution still needs accountable decisions and reliable records. Automating part of a process does not eliminate responsibility for the result. The commercial opportunity depends on whether the tool improves the complete workflow rather than merely producing a faster first draft.
TBJ’s examination of why AI agents need payment infrastructure addresses another piece of that implementation problem. A system capable of initiating transactions needs controls over authorization and spending. Technical capability becomes economically useful only when it fits the rules and operations of the business using it.
Demographics change the investment question
The study’s framework suggests that investors should look beyond a generic claim that AI solves labor shortages. Which workers are retiring? Which services face rising demand? Can the proposed technology address those specific tasks, and can the organization deploy it at an affordable cost?
Consider a hypothetical service provider facing a shortage of trained staff. Automating scheduling may free some time, but it may not remove the need for qualified people. The benefit could still be meaningful if it increases the time those people spend on their core work. That is a more realistic outcome than assuming an entire role disappears.
At the national level, training, participation and mobility can therefore remain important even when automation improves. Technology may change the mix of work rather than simply replace missing workers one for one. The adjustment can also create new demand for implementation, maintenance and supervision.
Productivity claims need a measurable denominator
A vendor’s demonstration that a task takes fewer minutes is useful evidence, but it is not yet a measure of economy-wide productivity. The full calculation includes adoption costs, quality, rework and the amount of the working day affected. A spectacular improvement in a narrow task can translate into a modest gain across an entire organization.
That does not undermine the case for AI investment. It makes the case more specific. Firms able to document sustained gains in a costly bottleneck have a stronger argument than firms relying on broad demographic narratives. The same standard helps customers compare a new tool with alternatives such as better conventional software or changes in staffing.
The BIS research adds a useful constraint to the current debate: ageing and automation do not line up uniformly. Its findings are an analytical framework, not a forecast that any particular company or technology will succeed. The next evidence to watch is deployment that improves output in the sectors where labor pressure is actually rising. That is where the promise of automation meets the structure of the real economy.
For advertising inquiries, please email . [email protected] or Telegram
