A new report claims that nearly a quarter of a million public sector jobs could be automated over the next 15 years, in a drive to slash costs and maximise efficiency via the smart application of new technologies, including AI.
Robots, says the report, will arrive like Uber in the UK’s public sector, creating an economy in which expert human citizens compete to offer their services at the lowest possible price while the robots run the machineries of central government, along with local functions such as health and education.
The report, ‘Work in Progress: Towards a Leaner, Smarter Public Sector Workforce’, has been produced by right-wing think tank Reform, whose recent speakers include Prime Minister Theresa May and Health Secretary Jeremy Hunt, and whose recent contributors have included former Prime Minister Tony Blair.
Reform sets out the context for its claims:
Current pressures mean the public-sector workforce must undergo radical change to deliver better value for money. Tight public spending means that public-sector productivity must break from its 20-year trend of near-zero growth.
However, it fails to mention the real-world background of central government budget cuts of 20-30 per cent, in some departments, while the Institute of Fiscal Studies reports that departmental spending overall has been slashed by 12.8 per cent since 2010-11.
Next, Reform sets out a smorgasbord of claimed future savings and organisational disruptions, saying that software robots and AI could replace up to 90 per cent of Whitehall administrators, as well as sweeping aside tens of thousands of jobs in the NHS, saving the Exchequer billions of pounds.
In this way, mass automation would deliver better value to taxpayers, says the think tank, with new leadership cultures ported from the private sector to speed its adoption. But this sets up an all-too-familiar scenario in right-wing political thinking: that private shareholder value and accountability are somehow interchangeable with public taxpayer value and accountability. They aren’t.
Counting the beans
The report’s list of apparently easy fixes to create public sector efficiencies include:
Any routine administrative roles have a 96 per cent chance of being automated by current technology. […] Central government departments could further reduce headcount by 131,962, saving £2.6 billion from the 2016-17 wage bill.
In the NHS… 91,208 of 112,726 administrator roles (outside of primary care) could be automated, reducing the wage bill by approximately £1.7 billion.
McKinsey estimates that 30 per cent of nurses’ activities could be automated, and a similar proportion for doctors in some specialities, enabling those skilled practitioners to focus on their non-automatable skills.
In primary care, a pioneering GP provider interviewed for this paper has a clinician-to-receptionist ratio of 5:1, suggesting a potential reduction of 24,000 roles across the NHS from the 2015 total.
In total, this would result in 248,860 administrative roles being replaced by technology.
Click-click-click goes the Reform abacus, as if the report itself has been written by an adding machine.
However, the ideological bean-counters at the think tank acknowledge that frontline roles, such as doctors, nurses, teachers, and police officers, require substantial expertise and personal interaction – skills that are less likely to be automated than routine administrative tasks. Yet there is still potential to use automation to increase productivity – and thereby to reduce headcount – it says. That much is true.
Senior management and “cognitive roles” will also be swept up by the march of the machines, claim the researchers:
These roles are least likely to be automated over the next ten to 20 years, but there are several areas where technology can improve senior officials’ work – increasing efficiency for them and the frontline staff who respond to their instructions.
The big picture
The big-picture context for this report is a raft of recent studies that present the employment impact of automation in near-apocalyptic terms. For example, consultancy McKinsey estimates that up to 45 per cent of activities in the US labour market could be automated using current technologies, while the Bank of England has claimed that up to 15 million jobs could be threatened by robots in the UK.
Last year, Dr Anders Sandberg of Oxford University’s Future of Humanity Institute predicted that in the near future 47 per cent of all jobs will be automated, adding, “If you can describe your job, then it can and will be automated.” It is in this context that Reform presents its checklist of (claimed) easy wins.
Some technologies will improve public-service delivery, adds Reform, citing the example of the various companies that are developing AI to diagnose medical conditions more accurately than humans, and in some cases to predict them. Certainly, IBM is among the providers using AI to predict the onset of health problems, but with the specific vision of augmenting human healthcare, not replacing it.
Next, Reform widens its focus to include government as a whole:
Whitehall should move from hierarchy to self-management, with teams organising themselves around tasks that need to be done. The Government Digital Service (GDS) has done this to great effect.
But this ignores the fact that the real-world GDS has been reduced from its once-core presence in Whitehall to a marginal group on the sidelines, while government procurement inches back to the old days of megabucks deals and big-ticket IT programmes – the culture that Francis Maude once sought to do away with entirely.
The real-world is messy and political, and nowhere near as binary and clean as Reform suggests. And let’s not forget that some of Maude’s ‘oligopoly’ tech providers are the same ones driving automation and AI. In other words, invite in mass automation and you’re inviting in the big technology suppliers, not slamming to door on them to save money.
Robots über alles!
Then the report veers into truly controversial territory, saying:
Public services can become the next Uber, using the ‘gig’ economy to employ locum doctors and supply teachers.
Flexible and temporary employment have been growing for decades, but the emergence of the gig economy, with workers supporting themselves through a variety of flexible jobs acquired on online platforms, has gained traction (and controversy) recently. ‘Contingent labour’ platforms – trialled in social care – may suit hospitals and schools as an alternative to traditional agency models.
It may also suit organisations who face seasonal peaks of demand, such as the need for HMRC to recruit additional capacity at the end of a tax year. 18F, the American version of GDS, has recruited coders for specific tasks by allowing them to bid for work at lower prices, in a reverse auction. Using such platforms in the public sector would show its commitment to delivering working practices fit for the twenty-first century.
This future vision of a generation of public sector workers – doctors and nurses among them – cut adrift from regular employment, workers’ rights, paid holidays, and more, undercutting their peers in a desperate attempt to secure a day’s ad hoc work is hardly a dream world for anyone except a bean counter. And an automated HMRC will be pursuing them for tax, presumably, vastly increasing the administrative burden for individual citizens.
That is a vision of citizens working for government, not government working for its citizens.
Set in the context of Brexit removing the UK from European human rights laws and employment regulations, and a UK economy whose belly has been slashed open to the wolves of the resurgent American right might be efficient, but also a nightmare for many citizens.
Then Reform turns its binary gaze to policing:
The UK should evaluate drones and facial-recognition technology as alternatives to current policing practice.
Autonomous crowd-monitoring drones could replace police-helicopter-operating roles by identifying issues and deploying police officers most effectively on the ground. Facial-recognition technology has been applied by police forces across the world, notably in the US and Israel.
Make way for robocop: applying the law as a set of simple, enforceable rules – in the context of the UK’s newly minted surveillance regime. What could possibly go wrong? [For more on this, see my diginomica report on the political algorithm]
Let’s be clear about one thing: automation, AI, and robotics will benefit human beings in countless ways, stripping away routine, repetitive tasks and augmenting human skills and experience. Many of these processes will be more efficient and cost-effective, and AI will also help us to uncover new data and new forms of research.
In most cases, AI’s vendors position the technology as a collaborative, augmenting force, not a blunt instrument to replace human beings.
And as I have said before, the problem in all applications of these technologies is that the first things that organisations automate aren’t routine functions, but their assumptions about the world – even if those assumptions are flawed, incorrect, or based on incomplete or inaccurate data.
Take the health service. An in-depth, investigative report for the NHS Support Federation published last month by researcher and journalist Nick Turner made the excellent point that cuts, efficiency drives, and political pressures have combined to remove data about the performance of the ambulance service from public view.
Underfunding and efficiency drives are combining to disable any comprehensive system of performance monitoring.
Put another way, the same push towards efficiency that is driving the rise of automation in government also forces essential data underground – data of the type that will help human beings test if the new technologies are actually working.
Like many think tanks, Reform applies hypothetical solutions within a set of idealised outcomes, revealing the kind of isolated, binary thinking that can be a deep-rooted flaw in AI research programmes themselves. [For more on this, see my recent Davos report for diginomica on the ethical challenges of AI].
These types of ‘solutions in a vacuum’ lack the messiness, political complexity, and self-interest of the real public sector, and ignore the real-world human outcomes that include essential monitoring and performance data being suppressed, hidden, or simply ignored.
While wide-reaching, the Reform report suffers from a surfeit of binary thinking, along with the implicit belief that all cost-savings are intrinsically wise and that automation is something that can simply be switched on in order for a result to be achieved.
In short, the report assumes that government can be made to work like a simple factory process: press a button and something is produced at the other end: Efficiency! Cost savings!
These were precisely the claims that were made for the offshore outsourcing sector a decade ago – just before many organisations kickstarted expensive repatriation programmes in the wake of disastrous customer feedback. In many cases, those assumptions about easy savings cost them dear, and since then many enterprises have used technology to become ever more remote from their users. Efficiency, it seems, is in the eyes of the beholder.
And this is an inherent problem with automation itself: when people propose it, they often think like machines, and only consider the outcomes within a set of idealised circumstances that barely resemble the real world. It never occurs to them to question their assumptions.
Take this from the Reform report:
The days of the topdown hierarchical organisation are slowly coming to an end.
And yet here we are with Trump in power, demonstrating just how wrong an assumption can be.
Image credit - Replacement Of Humans By Machines © ktsdesign - Fotolia.com