This is the third Part of four connected blogs on AI and the future of work. Part 1 focused on what AI is good for and identified some issues. Part 2 focused on the context of work in 2040. Here we discuss the future of work itself, and in Part 4 we use two different professions (IT and Healthcare) to illustrate how AI affects professions differently.

Definition of work

Work is defined an activity or job that a person uses physical or mental effort to do for money or as a volunteer (in this discussion we do not cover employment structures – see Part 2). Here we use the terminology of job roles to think about the future of work: a job role is the specific set of duties and responsibilities assigned to a person within an organisation.

Take up of AI in organisations

The World Economic Forum’s Future of Jobs Report stresses widespread expectations that advancements in technologies, particularly AI and information processing (86%); robotics and automation (58%); and energy generation, storage and distribution (41%), are expected to be transformative. These trends are expected to have a divergent effect on jobs, driving both the fastest-growing and fastest-declining roles.

The report also notes “… fuelling demand for technology-related skills, including AI and big data, networks and cybersecurity and technological literacy, which are anticipated to be the top three fastest growing skills.”

However we note that Pew Research in the US finds that 50% of Americans are more concerned than excited about the increased use of AI. So, which roles in organisations could be transformed by AI? In what way? We consider the characteristics of jobs that can be automated and separately those that can be augmented (for definitions see Part 1).

Example: job role people can relate to – auto mechanic

An auto mechanic’s work revolves around ensuring that a vehicle functions efficiently and is road worthy.

Their main duties and responsibilities include:

Repairs

• Diagnosing vehicle faults manually or with the aid of software and digital tools
• Preparing cost and time estimates for work to be done
• Repairing and/or ordering/replacing faulty vehicle parts
• Performing test drives to check repairs.

Maintenance

• Carrying out scheduled servicing and maintenance
• Updating vehicle service records
• Installing and fitting new vehicle accessories.

Stock control

• Maintaining a database of spare parts in the repair garage.

Which of these can be automated? Which benefit from AI augmentation? What is the differentiation?

By 2040, most cars on the road have significant on-board continuous monitoring and anticipation of failures, with warning messages routed to nominated service centres. The service centre automatically issues a recall notice to the registered driver if a physical visit is needed. Other problems are solved by software downloads – the driver will not even be aware when this happens - this is automation.

If a physical visit is needed, the auto mechanic uses the software and digital tools provided by the car supplier to diagnose the faulty part and the tools order a replacement. AI is augmenting the auto mechanic, providing prompts and information.

By 2040, cars without automated diagnosis are in the minority – their diagnosis and repair could be augmented by the use of more focused software tools, online instructions, and physically listening to or looking at the vehicle.

Cost and time estimates are automated by software platforms, as is the ordering process, while replacing faulty parts is augmented for physical components. Test drives are replaced by wind tunnel tests implemented automatically by the platform or, for driverless cars, a test route executed.

What could possibly go wrong? We are reminded of the incident in Project Hail Mary in which Grace, alone in space, asks his robot medical service for another pain killer. He gets the answer “not for another 3 hours”. So Grace resets the space capsule clock to Russian time - so that by the clock more than 4 hours have passed – and the painkillers are dispensed.

How will such “IT” support be provided for the cases in which automation has not solved the problem? Where will the skills lie? Who can “correct” the system?

Maintenance (preventing repair) is the main target of automation, relying on continuous monitoring to schedule routine servicing appointments. Many of the tasks in a routine service are physical – such as changing oil or replacing windscreen wipers. The expectation is that in a large main dealership, these are automated and performed by robots, but that the cost/benefit is not always there where routine maintenance is performed by the owner or a friendly neighbour.

Stock control is a well understood task, largely automated in warehouse-like surroundings. This is augmented by a system that advises the auto mechanic.

Differentiating between using AI to automate and to augment

Consider some sample tasks:

Using AI to automate (replacing) human capability

What tasks does AI complete better than people can? Tasks that require complex analysis such as:

• weather forecasting
• evaluating potential protein configurations for desired characteristics (AlphaFold and using to develop new medicines)
• analysing samples of newly discovered natural products in order to explore potential medical applications (e.g. new antibiotics).

What tasks does AI complete differently than people would, with a different outcome?

• Mowing a lawn, continuously and with smaller cuts each time so that cuttings mulch into the ground and do not need to be collected.

What tasks does AI complete as well as a human but at lower cost?

• Repetitive tasks with defined outcomes over 24 hour operation and several years e.g. factory assembly line tasks
• Analytical tasks such as contract analysis and comparison
• Management of complex networks like telecommunication networks.

Using AI to augment human capability

What tasks does AI augment that produce better outcomes for people?

• Continuous monitoring to flag anomalies to inform human decision making
• Analysing medical scans more quickly than humans in order to flag potential anomalies
• Scanning for relevant data sources.

The challenge is for AI to be used effectively to augment people’s capabilities, and for organisations to recognise the strengths of both AI and people. Not forgetting consideration of backup interventions as in Project Hail Mary.

In Part 4 we explore the use of AI in IT and Healthcare.