Algorithm-generated rotas: A good idea?

Using algorithms to allocate work has benefits for both employers and employees, but this should augment not replace the human touch

In the gig economy algorithms have been used to allocate work for some time. Take Uber and Deliveroo. For workers it offers a degree of autonomy – the opportunity to turn work down, choose when to work, and plan schedules and breaks better. For the employer it’s a good way of cutting down on admin and producing more compliant rotas. Such software is used widely in the recruitment world too; to match highly-skilled consultants to suitable roles.

However, increasingly HR teams are seeing the benefits of using algorithms to assign work to their ‘ordinary’ employees too.

Allocate’s HealthRoster is used by more than 60% of NHS trusts and more than one million staff. According to Allocate the health service is drawn to the financial benefits the solution offers, with its Collaborative Accelerator Programme reducing unbudgeted spend by £32 million each year across 18 NHS trusts.

Such software is also starting to be picked up by a diverse range of other sectors, with charity The Disabilities Trust and leisure company Junkyard Golf Club both using Planday. This Danish work allocation software uses factors such as workers’ roles, skills, employment status, working time preferences and cost to allocate shifts. The digital Punch Clock feature allows employees to check in on their smartphones to see any changes made to the schedule or when a shift begins. All employees’ hours are then automatically logged in the form of schedule reports that can be used to calculate pay and holiday.

“Using software to improve scheduling essentially means that you’re combatting the risks of human error having harmful effects on your business,” says Planday’s chief product and technology officer Chris Micklethwaite. “Rather than technology replacing humans, the software maximises efficiency and allows people to put their time into running better businesses.”

Giving staff constant oversight of their shifts can also help drive engagement among the workforce, feels Naomi Carey, director of HR at The Disabilities Trust. She points to staff appreciating the ability to see shift patterns and make appropriate changes – even when on the move. “The advantage of not being restricted to doing so at the office is key. They can sign in and out easily without having to keep a paper timesheet – a method of record-keeping that risks inaccuracies,” says Carey.

But not everyone thinks algorithms are the best way to allocate shifts. One problem, says CEO of workforce management software company HFX Nick Whiteley, is that “algorithms are often based on a false premise”.

“There is no such thing as a ‘perfect roster’ that can be mathematically modelled and achieved via algorithms,” he says. “They often ignore the local knowledge and human insight that is not codifiable within the system.”

Whiteley has developed several rostering solutions over the years (for NHS nurses and doctors, and for the airline industry) that use algorithms to design work patterns. “The algorithms on their own, while in theory may produce the perfect efficient roster, are not staff friendly,” he says. “What tends to happen is that, in real life, they are work patterns that people simply won’t work – often leading to industrial action.”

Mathematically there is often nothing wrong with the algorithms, he explains, and they have been used to great effect in sectors such as retail. “The problem is the underlying assumption that there is a perfect roster,” he reiterates. “Even with today’s computer power you can’t go down every single path. Algorithms don’t take into account things like local knowledge of the operations manager, how much time off could affect things like training capability, and detailed information about how managers work with different teams.”

They can be mathematically “perfect” but they’re not always workable, Whiteley warns: “Moreover, changing someone’s roster can be even more dramatic than changing someone’s pay.” Computers can do a huge amount of grunt work but what they can’t provide “is human insight.”

Rather than replace individual rostering, algorithms should augment the responsibilities of employees, says Cyrus Gilbert-Rolfe, MD EMEA of communications platform SocialChorus. “Scientific understanding of how different employees work best is inevitable, right down to the individual level,” he says. “That is part of how AI is continuing to transform the workplace.”

It’s well known that AI and automation are eating into tasks that would traditionally have been done by humans. Gilbert-Rolfe says: “If we think of the whole spectrum of human work activity, then one end is the hardest thing we ever do and the other is the simplest thing we ever do. One end might be understanding and managing another person and the other end is repeatedly pressing a button. AI is eating into the end of the spectrum that starts with the button pressing.”

Yet the shift towards a scientific method of people management seems to be mounting. “At the moment we still think of work as something we do during a fixed period of time, at a fixed location, most days,” says Gilbert-Rolfe. “That idea is undergoing severe disruption in things like remote working and explosion of the freelance and gig economies, but the underlying idea is still present.”

Gilbert-Rolfe believes that in the near future this mindset is likely to shift or be shifted (as in the case of Uber drivers). “More commonly tasks are assigned to different people based on skills or preferences they have, either stated or perceived by managers. It is extremely likely that more tools will become available – both to employees and managers – that measure things like behaviour, physical responses and attitude and make recommendations for the best tasks, best times of day, the best make-up of teams, and so on.”

So is such a shift a bad thing? “It is very hard to see how [it could be] if we assume that the people writing the algorithms know what they are doing,” says Gilbert-Rolfe. “But it is a change with increasing momentum. And like most change it is somewhere between exciting and terrifying.”

This piece appeared in the July – August 2019 issue. Subscribe today to have all our latest articles delivered right to your desk