Introduction

This is the last of four connected blogs on AI and the future of work. Part 1 defined the terminology, discussed what AI is good for and identified some issues. Part 2 focused on the context of work in 2040. Part 3 considered the future of work in the large, and in this part we use roles in two different sectors (IT and Healthcare) to illustrate potential uses of AI.

The scope for AI in healthcare is of wide interest. Actual healthcare systems have been studied in a number of geographies. Many studies are in hospital settings, where processes are well defined so it is possible to measure times for each task and assess the quality of outcomes.

In IT, the impact of AI on coding and related development tasks like documentation has been studied. There are analyses of areas in which AI could contribute to operational roles, but we have not found any data on the effective use of AI in operational roles – managing complex IT systems. So, the analysis below of an availability service manager’s role is qualitative, based on work by the BCS Availability Working Group and their sources.

Radiologists in the age of AI

Nine years ago, Geoffrey Hinton declared that we should stop training radiologists, because in five to ten years, they would be replaced by AI.

But, as Noah Smith observes, in 2025, American diagnostic radiology residency training programs offered a record 1,208 positions across all radiology specialties, and the field’s vacancy rates are at all-time highs. In 2025, radiology was the second-highest-paid medical specialty, with an average income of $520,000, an increase of about a half since 2015.

One of the reasons for the unrealistic projection is that the AI model makers did not know what radiologists actually do. They know that radiologists read scans, but only about a third of radiologists’ time (according to a recent study) is interpreting images. This can be automated, and radiologists also spent a considerable time in teaching, another candidate for automation.

Among the other activities of radiologists, some are suitable for augmentation: for instance, a further tenth of radiologists’ time was in image guided procedures, and consulting with other staff took nearly a quarter of their time.

Both automation and augmentation can contribute to productivity increases. But increases in productivity reduce cost. This will increase the demand for radiologists: the increase in demand can outweigh the increase in efficiency, a phenomenon known as 'Jevons Paradox'.

This has historical precedent: in the early 2000’s hospitals swapped film jackets for digital systems. Hospitals that digitised found that time to read an individual scan went down. A study at Vancouver General found that the switch boosted radiologist productivity by a quarter for plain radiography and by double for CT scans, within a year of going filmless. However, the overall American utilisation rate per 1,000 insured patients for all imaging increased by nearly two-thirds from 2000 to 2008. An average visit to a hospital was associated with more imaging.

Hinton might someday be right, but as of right now he is wrong.

IT Availability Service Manager In The Age Of AI

An availability manager oversees the end-to-end availability of the organisation’s services: the role coordinates across the organisation and out into the supply chain and to users. This role is often event driven, and covers a wide variety of tasks – like jobs in many other domains.

Tasks that can be automated include many of the support functions such as: system mapping and documentation; updates of new versions of software; mapping the supply chain; managing contracts and Service Level Agreements (SLAs); testing – the generation of test data from the documentation and scheduling tests, analysing results and reporting exceptions; business continuity planning i.e. identifying important business functions and implementing availability requirements and disaster recovery plans; monitoring and incident detection; and generation of response options.

Many of the management tasks are augmented by the data collected and analysed as above, for instance:

• Controlling and managing service availability to deliver agreed levels of availability in a cost-effective manner; identifying and managing risks to availability.
• Continuous improvement, updating availability plans and documenting progress on improvement actions; addressing instances of unavailability through corrective and preventive actions; change assessment – planning with teams for upcoming software or hardware changes; and risk analysis.
• Stakeholder communications – communicating systems status and any major issues to senior management; liaising with suppliers; warning users of any planned outages or improvements.

Conclusions

Parts 1 and 2 identified the qualitative difference between automation and augmentation, and explored some of the ways in which the context for work in 2040 will be different from today, with a context of competing political and value systems hindering tackling of existential and global challenges. Technology capabilities may advance faster than society can adopt and use them. Social Media platforms of today are relics. Small family sizes and urbanisation create new demands for communication, co-operation and essential services.

In Part 3 we used the example of an auto mechanic to illustrate the challenge for AI: to effectively automate, and also to augment people’s capability; and for organisations to recognise the strengths of both AI and people. We then explored the characteristics of tasks that AI does better than humans, those AI does differently and those it does as well as humans, but at a lower cost.

In this, Part 4, we have used the examples of jobs in two different sectors to illustrate potential use of AI automation and augmentation. The extent and nature of each depends on the characteristics of each task and the context – global, organisational or personal. Currently, most organisations are in pilot phases with AI, so that any data is often based on small studies: by 2040 the picture will be clearer!

There is legitimate concern across many professions overthe decline of human cognitive skills due to the use of AI. Is this just concern over introducing new technology or is it something more fundamental?

In conclusion:

• AI is neutral
• Automation and augmentation are qualitatively different and need to be used appropriately.

We hope that our analysis can contribute to deeper understanding of how AI can be used for the benefit of people and society.