Despite predictions of the death of the office in the 1990s, remote working has been slow to take off. Across the EU the share of the population working from home has hovered between four and five per cent for the past two decades.
However, Covid-19 looks likely to change all of that. In 2015, researchers at Stanford University found that remote work increased performance by 13 per cent due to fewer breaks, sick-days and a quieter working environment. And several anecdotal studies of remote working during the pandemic show that people working from home have become more productive.
In 2021, however, we will have to grapple with its downsides. While working from home brings efficiency gains in the short run, the danger is that it will imperil the innovation that drives business performance over the long run. Indeed, efficiency is the enemy of innovation, which is fundamentally about exploration. Having everyone working by themselves makes it hard for people to interact and explore new possibilities.
The solution to this dilemma will come from artificial intelligence (AI). The inherent trade-off between exploration and efficiency is well known to AI researchers. One question that those working in AI often have to grapple with is how often an algorithm should take actions that it hasn’t tried, as against actions it has already tried that will usually lead to some reward.
Untried actions can yield spectacular results. For example, when the DeepMind computer program AlphaGo beat Go world champion Lee Sedol in 2016, it did so by exploring moves most human players had never seen before. Prior to move 37 in the second match against Sedol, AlphaGo had calculated that there was a one-in-ten-thousand chance that a human player would make that same move. And the adventurous gamble paid off.
Human innovation involves a similar process of exploration and, to facilitate innovation, companies must get their employees to “collide”. Before the pandemic, this was achieved through open-plan architecture that encouraged “water-cooler” moments of unplanned encounters. But, with many employees working from home, corporations will have to find different ways to facilitate these kinds of random interactions.
The prime reason why, until now, people prefer to work together in person rather than online is that digital technologies have provided poor substitutes for the sporadic encounters that happen at work. But AI has the potential to change that. AI is already good at matching, whether it is finding the right film on Netflix or the right partner on a dating app.
In 2021, companies will be throwing resources at developing this kind of matching AI in the workplace. Based on employees’ emails, Google searches and other data, AI algorithms will be able to deduce what people are working on and their current interests and will act upon that by making digital introductions that would otherwise not have happened. Employees will then evaluate the usefulness of each digital encounter, providing feedback for the AI to learn from.
As more companies grapple with the problem of powering innovation at a time when many are forced to work from home, we will see more AI applications being developed to promote sporadic digital encounters in 2021. If we can get this right, as the economist Frances Cairncross observed back in the 1990s, the world in which millions of people trooped from their home to the office each morning, and reversed the procedure each evening, may finally strike us as bizarre.
Carl Benedikt Frey is director of the Future of Work at Oxford University’s Martin School and author of The Technology Trap (Princeton)