Manufacturing must leverage AI to get the most from human touch

21 November 2018 5 min. read

While AI is often spoken of as being a phenomenon that will be bad for jobs in the manufacturing sector, it is becoming increasingly apparent that the human touch will remain an essential asset to the industry. Instead, AI should be wielded to improve the performance of the human workforce, rather than replace it, according to a new study.

The continued march of innovative new Artificial Intelligence (AI) technologies have brought with them a host of debates surrounding their implementation. One of the most consistent discussions on the issue is the extent to which automation, machine learning and AI may lead to job losses

While it is still largely impossible to know exactly how much AI will impact employment over the coming decades, the agreement is almost universal that the tech's implementation will see manual labour and manufacturing jobs decline. To this end, the UK has already seen the number of manufacturing jobs fall by 17% over the past decade, with around 620,000 jobs lost, according to estimates based on data from the Office of National Statistics, and much of this was said to be the result of automation.

However, according to a new study by consulting firm A.T. Kearney and AI start-up Drishti, while robots have made inroads on the factory floor, humans are still required for the vast majority of manufacturing tasks. The study analysed the results of one-on-one interviews with senior-level manufacturing employees and a survey of more than 100 executives across a variety of industries, including automotive, metals, aerospace, transportation, plastics, and food and beverage. In the process, researchers found that for now at least, human labour remains essential to the industry.

Importance of data sources

More than 72% of factory tasks are performed by humans, while humans still create an almost equal proportion of value, at 71%. According to the paper, this will see humans remain integral to the manufacturing process for years to come, as they are significantly easier to derive profit from than machines. Robots require significant capital, skilled resources, long installation periods, and attendant programmers and engineers. They cannot be attributed a lower amount of resources, or they cease running. At the same time, somewhat brutally, humans can be paid less than the value of their labour, and find ways to continue functioning.

The report itself added that this also sees the human touch boost creativity, something essential when creating products which can stand out from an ever more crowded market. The researchers wrote, “We can adapt. We are dexterous beyond anything robots are capable of today. And as manufacturers increasingly strive for lot sizes of one to satisfy customer demands for personalization and customisation, these qualities are more valued than ever before.”

At the same time, however, this human element comes with its own degree of risk. To be more precise, it is significantly less consistent than its mechanical counterpart. Survey respondents said that 73% of the variability on the factory floor comes from human workers, not machines. At the same time, a similar 68% of defects were said by the executives to be caused by humans. This suggests that in order to get the best of both worlds, a “centaur” strategy of hybridisation is required, in which AI can be leveraged to source anomalies in the human work and improve upon them.

Operations decisions that are influenced by human factory analytics

To this end, however, the Drishti-A.T. Kearney study found that management was slow to adapt. Citing estimates from the International Federation of Robotics and Goldman Sachs that some 1.7 million robots work alongside the world’s 345 million factory workers, the paper suggested that as of yet the ways in which data is leveraged to manage human labour in manufacturing is limited. A majority of those polled did not feel that leveraging data on poka yoke (systems for avoiding human mistakes) was very or fairly important.

The authors, regardless, state that senior-level engineers spent more than one-third of their working hours manually gathering data about worker productivity and that humans continue to perform the majority of critical manufacturing tasks. Survey respondents said that data on human tasks harvested through these methods heavily or fairly influenced initiatives such as daily staffing (78%); workforce management tasks such as hiring and training (80%); capacity planning (79%); job quotes (77%); process engineering (71%); and identifying automation opportunities (70%).

Commenting on the findings, Prasad Akella, Drishti’s founder and CEO, said, “There isn’t a systematic process for gathering statistics. I can tell you that it’s a pervasive problem in manufacturing based on the numerous meetings I have had with executives in manufacturing. Just about every company I talk to encounters the same lack of visibility into tasks performed by human workers.”