worker in apron, cap with tablet, at production line in factory
ATD Blog

Role of Artificial Intelligence in the Manufacturing Skills Gap

Wednesday, September 19, 2018

Two million jobs are expected to go unfilled between 2015 and 2025 due to the skills gap in manufacturing, according to Deloitte and the Manufacturing Institute. That is an astounding statistic, and from my own interactions with manufacturing leaders it is driving real concern. In my interactions, the workforce and the skills gap come up often and with tremendous trepidation. This is true for essentially all businesses, from small to large, and in any industry. There are many influencing factors, such as a large generation of retiring workers, a younger generation that shows less interest in manufacturing, and a greater focus on advanced technology in manufacturing. The question that we need to answer is, Can we get to the root of these concerns for manufacturers and address them there?

Taiichi Ono is widely regarded as one of the great thinkers and leaders of modern manufacturing. He was one of the driving forces behind Toyota’s production system (Lean). He is often credited with the methodology called “the five whys.” The premise for the five whys is simple: Ask why at least five times to get to the root of a problem. The following example shows how this would unfold with regard to the skills gap:

1. Why are leaders concerned about the skills gap?

  • Because many skilled workers are required to meet cost and production targets.

2. Why are many skilled workers required to meet targets?

  • Because unskilled workers are not as productive as skilled ones.

3. Why are unskilled workers not as productive?

  • Because unskilled workers lack the knowledge that a skilled worker has.

4. Why does an unskilled worker lack knowledge?

  • Because they must gain it via on-the-job experience.

5. Why is the knowledge available only from experience?

  • Because it is too specific with too many variables to be condensed, arranged, and taught outside of the production environment.

If we agree that a primary reason for a skills gap is too much complexity and nuance to train in any way other than on-the-job experience, then the question becomes: Can we use technology to address the issue of needing experience to gain the specific knowledge of complex interactions? The answer to that question is yes—artificial intelligence (AI) is well suited for this task. To be clear, I’m referring to AI as the application of learning algorithms. These learning algorithms are capable of taking large amounts of data and finding patterns and correlations that can be used to estimate or predict outcomes. This is the process of creating models. Humans have been creating models intuitively and intentionally forever. But, only a few have ever formalized those models in a way that can be shared with others. AI gives us a standardized approach for automatically creating, sharing, and implementing those models many times over.

A good example of how AI can be applied in manufacturing is product quality. Ensuring that quality products are made is a key responsibility of manufacturing personnel. In many cases, there is a significant amount of skill necessary to interpret the results of a product’s quality test and know how to adjust the manufacturing process to ensure quality standards are consistently met. Fundamentals for making this decision are taught in post-secondary education for manufacturing disciplines, but the nuances for a particular product are learned on the job. Assuming that the quality data and machine data are available in digital form and able to be linked together, then a set of algorithms can be assembled to enable the artificial “brain” to “learn” the links between operator actions and product outcomes. When this linkage is learned, the resulting model can be used to guide operators to take the appropriate action for the current situation. This guidance will enable experienced operators to make a faster decision and inexperienced operators to make faster and more accurate decisions the first time.


Operations planning and scheduling is another example. Today this is typically done in localized pockets without real-time feedback from the greater enterprise. The planning models are generally complex and created on an ad hoc basis. This is because there are numerous complex interactions that a person does not grasp until they have gained experience dealing with the subtleties of their process. In a similar way as the quality example, assuming that inputs and outputs are measured and can be linked, then the artificial brain can learn how changes in sequencing and flow through the manufacturing process can affect the overall production times. These models can account for real-time information and provide recommendations for the higher levels within the manufacturing business’s organization to make better decisions.

When AI is used in these ways, it is not replacing a person. It is augmenting their current knowledge or cognitive processing abilities. The phrases “augmented decisions” or “decision augmentation” describe this usage. A human is still in the loop to make the final decision and take the action, but augmenting their decisions will allow them to make faster decisions with fewer errors. People will be more productive and will enable their business to meet its targets. It should also be stated that when manufacturing operations are more productive, businesses will grow and manufacturing will expand, and this will drive the need for more manufacturing workers. Therefore, the takeaway should not be that technology will solve the problem, but that technology will play a role. An appropriate strategy is to continue to include all the traditional recruiting and development coupled with the adoption of technology.

About the Author

Andy Henderson is vice president of engineering at Praemo. Praemo delivers productivity-improving insights from manufacturer data using artificial intelligence analysis tools. Prior to joining Praemo, he earned a bachelor’s degree in mechanical engineering from Georgia Tech and a PhD in automotive engineering from Clemson University. He has worked at Caterpillar and two different divisions of GE, including a period as the industry solutions director for heavy industry manufacturing within GE Digital. He stays active in the manufacturing community at large through the Society of Manufacturing Engineers (SME) and was a recipient of SME’s Outstanding Young Manufacturing Engineer award in 2016. He has stayed actively engaged in manufacturing research at Clemson for most of the past decade.

Be the first to comment
Sign In to Post a Comment
Sorry! Something went wrong on our end. Please try again later.