How can organizations 'lay out the pipes' for automation success?

Over the past few years many have predicted that “the robots” will take over the jobs of millions of workers. Some predict even grimmer days ahead, claiming that the dystopic, jobless future will generate no productivity gains whatsoever. And when it rains, it pours: the pandemic-related disruptions of work and companies’ balance sheets have only intensified these fears. According to the World Economic Forum, 80 per cent of business leaders are accelerating the automation of their work processes, and 50 per cent have stated that they are ready to further speed up the automation of jobs in their companies.1

The key to separating fact from fear-mongering is to consider the millions of workers at the centre of this debate. Will they lose their jobs, see a drop in their salaries, or will their levels of productivity and well-being actually be boosted as a result of human-machine collaboration?

Analysing automation data to understand its implications

The potentially adverse effects of automation, such as reduced salaries and job losses, are, of course, one possibility. Yet there are also real upsides to using automation technology, as has been the case for any other workplace equipment.Think of it this way: calculators did not replace mathematics teachers. They also did not reduce their salary. Instead, it made teachers’ jobs less mundane by automating calculations while increasing time resources that could be devoted to imparting knowledge and honing students’ skills. Hence, data-driven analysis will help us better understand the impact of automation on workers.

Esther Duflo, MIT economist and the winner of the 2019 Nobel Prize for economics, argues in a 2017 essay that “economists should seriously engage with plumbing”.2 What Duflo means by this is that economists should pay close attention to what is actually happening on the ground. And that they should design policies accordingly to make economics—or automation—work.

The first task will be to “lay out the pipes”. But how can we be sure that this task is being carried out properly? We argue that automation data needs to be analysed in a more experimental and detailed way. Ideally, such an analysis should go beyond the forensic tracing of links between productivity, headcounts and pay. A more detailed approach would also shed light on a variety of potential effects, including the distribution of automation costs; benefits for workers with different skills at various stages of their careers; the creation of new tasks suitable for human-machine collaboration, and how performance appraisals and salaries might change due to automation.

RPA in real-time: a data-driven case study

To identify these relationships, we reviewed the impact of the introduction of a specific form of automation, Robotic Process Automation (RPA), on a typical workflow, and conducted a study that involved around 500 employees.

We focused on the impact of automation on customer support in IT services. The 500 employees included in our study were at different stages of their careers and were responsible for different tasks and activities. Automation was introduced for tasks such as resolving certain client requests, for example, system resets. It was also used to anticipate potential problems or requests before they could arise by using system monitoring. On average, there were around 30,000 such interactions per month.

Productivity gains and pay increases

We tracked the workers over a period of two years: one year before the introduction of automation, and one year after. We recorded productivity metrics (e.g. how long it took to resolve requests) and merged them with data such as worker demographics, annual performance assessments, pay and time spent “on-the-job”. This allowed us to compare outcomes before and after the introduction of RPA.

We found that RPA generated significant productivity gains (figure below).

For example, the average time it took to resolve a request (“ticket”) decreased by one-fifth in the year following its implementation. This was accompanied by an impressive 30 per cent decline in solution errors (figure below), which are requests that require a second look because the initial solution did not resolve the underlying problem.

Interestingly, these productivity gains were shared by all stakeholders, including the workers. Nearly 95 per cent of employees were still working for the company one year after RPA implementation. This includes 12 per cent of workers who moved to other business units. Those who moved to other units benefitted from company retraining and career development programmes. The upshot was that the normal rates of worker turnover remained unaffected by RPA.

In addition, pay raises in the year following RPA rollout were up by 3 percentage points for those who remained in their original business unit compared to workers who were redeployed to other business units.

Our results also show why it is important to keep an eye out for side effects. In our study, the lower pay raises among employees who had been redeployed to other business units reflected their performance in their previous role. This implies that organizations must create opportunities for employees who do not gain from the benefits of automation to help them stay engaged and realize their full potential.

Understanding the potential consequences of automation can help increase its benefits. And the optimal use of internal value levers—such as talent or data—can make all the difference in this respect.

Adoption must be accompanied by agility, which requires taking account of case specifics. Agility can be achieved, for example, by providing customized training for employees engaged in a new project to ensure a smooth transition (figures below).

Finally, the impact of automation on employees depends significantly on the tasks being automated. There is higher demand for employees at more junior levels when RPA changes the workflow away from urgent, high priority tasks by providing the system with more stability, as was the case in our study. But the opposite can be true if automation replaces less urgent, low priority tasks.

Preparing businesses for automation

Concerns about automation’s potential negative effects are not supported by the data we analysed. It is true that companies that do not approach automation in a strategic manner might face a thorny path. But these instances are avoidable. Proper measurement and data analysis to better understand the implications of automation implementation, the monitoring of its impact and making necessary course corrections as needed, for example, by providing timely training and information to sustain workers’ engagement, are just a few options of how both companies and the workforce can thrive alongside automation.

More studies like the one we conducted are necessary to gain more insights into the real-life impacts of automation on the workplace. Although human-machine collaboration can benefit companies and their workers, there is so much more we can learn to make automation work for our people, not against them.