Will AI steal our jobs? Wrong question. AI performs tasks and most jobs involve a combination of tasks. The AI impact depends on organisational choices, not just the technology. David Buchanan and Steve Macaulay explore how the outcome for jobs depends on a combination of replacement, compensatory, and augmentation effects.
AI may steal bits of your job, but very few jobs will be completely stolen. Simple admin, content generation, customer support, data entry, financial analysis, and manufacturing tasks can be automated. But this gives staff time to handle more complex and interesting tasks.
New technologies actually create new jobs
Past experience with new technologies has taught us two lessons: choice in how different aspects of work come together, and that new technologies actually create new jobs.
First, we can choose how work processes are designed, and in particular how the social and technical aspects of an organisation work with each other. These are called socio-technical system (STS) design choices.
These choices are not determined solely by the technology and its capabilities, but they will reshape work and affect job numbers. ‘Stealing’ jobs, therefore, is the wrong metaphor.
Second, new technologies create new jobs. However, these are likely to demand new combinations of skill and knowledge. The role of HR/L&D is therefore central.
Replacement, compensation, augmentation
To understand how any new technology will affect work and employment, we have to consider replacement effects, compensatory mechanisms, and augmentation processes:
1) Replacement effects
The current debate focuses on this issue. For example, robots take over manufacturing tasks. Agentic AI takes over administrative tasks. There is concern that many entry-level jobs will be automated. But if that happens, where will companies find experienced recruits to fill their next-level positions?
In 2023, Elon Musk made a prediction, that artificial intelligence will eliminate the need for all jobs. In the 1930s, the economist John Maynard Keynes made a similar prediction about the impact of automation, and we know how that turned out. But is AI different? Will the impact be greater this time?
2) Compensatory mechanisms
There are many mechanisms which will compensate for or mitigate replacement effects.
- New products and services need new infrastructure (shops, factories, offices, distribution chains), which create jobs in those areas
- Reduced costs from innovation lead to lower prices which increase demand for other goods and services, creating more jobs
- New technologies are not always implemented rapidly. It takes time to solve technical and organisational problems and scrapping existing facilities can be costly
The benefits of new technologies may not be clear. Organisations experiment with new systems gradually, to hedge the risks. Investment in new technologies is based on the expectation that the organisation’s market will expand, in which case the existing workforce may be retained, or increased.
Autonomous vehicles may replace delivery drivers. But it will take time to replace the fleet of conventional vehicles, and the self-driving kind still have to be loaded, programmed with their destinations, unloaded, cleaned, maintained, and serviced. They will also have to be manufactured and distributed to customers in the first place. The net effect on job numbers is thus difficult to predict, so could turn out to be neutral or positive.
And just because a task can be automated, doesn’t mean that it should be. There will always be situations where we need a human being in the loop, to bring experience, empathy, judgement, and oversight.
3) Augmentation processes
Less widely discussed are ways in which AI can augment or complement human abilities. In medicine, AI systems help doctors to diagnose and treat complex conditions. AI can analyse records of patients with similar conditions and suggest diagnoses. Doctors then interpret the results and make the final decisions about treatment.
Virtual assistants like Siri and Alexa support our decision making by providing information on demand. For the organisation, AI can analyse market trends, consumer behaviour, and competitor data, and identify business opportunities and risks. But management still decide the corporate vision and strategy based on cultural, financial, and other considerations.
There are of course risks. Those hit by replacement effects may not have the skills to move into new roles. Change may happen too fast for people to catch up. Unlike traditional automation that replaced manual labour, AI threatens knowledge workers. Positive outcomes are not guaranteed, emphasising the importance of strong HR input.
The radiologists’ story
Radiologists give us a good example of augmentation processes. In 2016, an AI expert predicted that AI would make radiologists redundant in five years, by automating the interpretation of disease patterns in X-rays and other medical images. In 2017, an AI system designed to detect pneumonia outperformed highly experienced radiologists. These systems are now in widespread use.
However, the number of radiologists employed in the UK rose by 40% between 2016 and 2025. This is because AI tools are used by radiologists, not instead of them.
Humans and AI systems both make errors which can have serious consequences. Working together means more accurate results. In a survey of clinical directors by the Royal College of Radiologists, 40% said that AI had increased their workload, only 6% said that it had been reduced, and the rest saw no change.
There are several reasons for this. There are problems integrating AI with existing IT systems. AI systems create new tasks such as monitoring accuracy. Speeding up parts of the workflow creates bottlenecks elsewhere. AI-assisted MRI scans are performed faster, so radiologists have more images to review.
The datasets for training AI do not cover all types of patient problems. A system trained on adult X-rays, for example, is not reliable for children. A human radiologist can identify problems in an image that an AI system has not been trained to look for. Radiologists also decide what imaging is required, and use their judgement, experience, and knowledge of a patient’s history to diagnose and monitor treatment in collaboration with other specialists.
That AI expert suggested in 2016 that we should stop training radiologists because they were not needed. Health services today worry about shortages of radiologists. We can expect to see similar outcomes in other occupations.
What is your people strategy?
The impact of AI on employment depends on the combination of replacement effects, compensatory mechanisms, and augmentation processes. The outcome does not depend on the technology, but on how you use it. You can replace people. You can experiment with novel ways developing new products and services. You can use it to augment the skill and knowledge of your staff.
Socio-technical system design choices affect key HR and L&D policies: work design, skills and competencies, reskilling and upskilling, technical and corporate knowledge, training and development, career management, compensation, performance appraisal.
Change management is another critical factor. Expect resistance if you simply impose new technology. Invite participation in how best to use it, on the other hand, and you may be surprised by a creative and welcoming response.
Debate in this area is often technology-led, all about what the kit can do. Those who design and make the stuff tend to exaggerate when it comes to capabilities and benefits and, therefore, HR and L&D policies are in danger of becoming add-ons, after the fact.
It’s much better to get ahead of the game, where HR and L&D considerations can head the list when it comes to decisions about when and how AI will be implemented, used, and developed.
Five priorities for L&D leaders
In guiding AI implementation, these are priorities for L&D:
- Be involved in technology application decisions from the start
- Build the business case for socio-technical work system design, meeting technological, organisational, and human needs
- Ensure employee involvement in the socio-technical design process is critical; they are the ones who will be working with the new technology
- Determine which tasks to automate, and which require human input and oversight
- Launch reskilling/upskilling programmes early
Be involved in shaping the future
AI will reshape the world of work. But it will not have a simple ‘robots steal jobs’ impact as often portrayed. What matters is how we redesign work, redeploy skills, and integrate human and machine capabilities.
This is no time to leave it to the tech bros. These are HR/L&D decisions, which will determine whether AI is disruptive, or a catalyst for better work, and improved organisational resilience and performance.
A jobs apocalypse is unlikely. A jobs bonanza is a possibility. The reality probably lies somewhere in between. But that depends on decisions about how we use AI.
Steve Macaulay is an Associate at Cranfield Executive Development. He can be contacted at: s.macaulay@cranfield.ac.uk
David Buchanan is Emeritus Professor of Organisational Behaviour at Cranfield University School of Management. He can be contacted at: david.buchanan@cranfield.ac.uk
