Even though AI has become one of the hottest topics in manufacturing today, most manufacturers are at the start of the adoption curve. To progress, companies need to embrace an evolutionary implementation strategy.    By Thomas Leeson

It’s always tempting to begin an article on AI with some form of science fiction analogy, but the truth is that the technology has been around almost as long as the genre! Some of the people reading this, like me, may remember the burst of excitement around AI in the 1980s and 90s. We’ve come a long way since then, and many facets of AI, especially machine learning, have become very mature.

The increasing digital transformation happening within manufacturing is bringing the potential of AI into focus. International Data Corp., a technology research firm based in Framingham, MA, suggests that manufacturing companies are “at the heart of a perfect storm, both living with and seeking to exploit disruptive technologies such as cloud, big data, AI-assisted analytics and the Internet of Things (IoT), while facing increasing IT security challenges, regulatory pressures and a changing workforce”.1

The explosion of big data and IoT is pivotal. Taking IoT as an example, it is estimated that there will be upwards of 75 billion connected devices in operation by 20252, the majority of which will be in the manufacturing sector. When you consider that a single device can produce gigabytes of information each minute, the volume and velocity of data within production and operations can easily become out of control – or, more often, simply ignored.

While 60% of executives that McKinsey surveyed said that IoT data yielded significant insights, 54% admitted that they used less than 10% of that IoT information3. Manufacturers are faced with a virtual tsunami of data. And IoT information is only one aspect of a new digital ecosystem of systems, people, and things.

The barriers between operational technology and information technology are breaking down as companies look towards becoming data-driven in all aspects of business from product design to smart factories to supply chain visibility to enhanced customer experience.

The Role of AI 

Making sense of that data is where AI comes in. In combination with advanced analytics, it can bring together information from a wide variety of data sources, identify patterns and trends, and provide recommendations for future actions. It provides the basis for new levels of production and business process automation as well as improved support for employees in their daily roles and in decision-making.

It’s clear why a recent Forbes survey found that 44% of respondents classified AI as “highly important” to their manufacturing function in the next five years, while 54% went so far as to say it was “absolutely critical to success”4. Why, then, did the Manufacturing Leadership Council’s Factories of the Future survey5 show that only 8% of companies have currently started implementing it?

Some time back, McKinsey coined the term “data trilogy” to reflect the emergence of three technologies – big data, analytics, and AI –that were disruptive in themselves but together had the potential to completely transform business. It can be loosely said that they followed each other. Of course, analytics and AI both existed before big data but big data has been one of the driving forces behind the adoption of advanced analytics as analytics is now an important driver in establishing the need for AI.

The adoption curves of each of these inter-linking technologies are getting flatter and faster. For example, the Manufacturing Leadership survey found that, while only 8% were currently using AI, a further 50% expected to deploy it within two years.

If AI is still nascent in manufacturing today, these results suggest it could become mainstream in under 24 months. This speed of adoption can be achieved because many of the companies reporting success with AI today – 54% say AI has already increased productivity6 – have done so by integrating AI into their existing big data and analytics solutions. In its predictions for 2019, PwC found that integrating analytics and AI is the top priority for organizations7.

A Steel Maker’s Experience 

One company, a steel manufacturer, has been rated a leader in its industry sector for well over a decade for having the safest plant, highest quality, best on-time delivery, and customer service. It’s achieved this position through a focus on delivering the most value from its data. The steel maker began by introducing analytics to gain a clear picture of its operations and is now beginning the process of investigating where AI can deliver business improvement.

The company’s IT solutions manager says analytics have helped “clean up our data. I’m now looking at it and how I can benefit from the data that’s there. I can really see what I’m doing” and AI will take this a step further to understand “what’s coming in the future? How do I look for things? Where do I need to improve?”

The barriers between operational and information technologies are breaking down as companies try to become data driven. This is where AI comes in.

He believes that augmenting the company’s analytics capabilities is the best way to implement AI into the organization. “It’s all about known entities, understanding what’s happening and what the result is,” he says. “By adding AI to our analytics, we can narrow down eight or 10 potential approaches to, maybe, one or two so that we can make faster, smarter decisions.

“Our experience so far has been that it should be straightforward to add AI into our analytics systems. It becomes a business rather than a technical decision how to proceed. Where can we deliver the most benefit to the company?”

Key Applications for AI 

Taking an evolutionary approach to AI adoption can speed the time to value of your AI program while reducing the cost and complexity of the process. The question becomes what element of your business lends itself to the capabilities of AI and will return the most value quickly?

The steel company manager identifies three key areas where AI can transform manufacturing: “I’d say that sales and business development, the supply chain, and the plant is where we’ll see AI really flourish. From our perspective, we have invested over many years to make sure that we’re operating at really high levels in the plant so, although there are incremental improvements we can make here, sales and the supply chain are likely to be where we can realize the biggest benefits.”

Costing and sales development were two of the first areas where the steel company introduced analytics. It wanted to know exactly how events such as plant delays and bottlenecks affected profitability as well as identifying the blends and mixes of steel that would deliver the highest margins. Analytics has delivered a better understanding of customer buying trends and needs and AI offers the opportunity to build upon this.

“Steel coils are quite a niche product so there are only so many potential customers for us throughout the world,” the IT solutions manager says. “We have to be sure that we are producing the value-added products and services that our customers require. We need to quickly and accurately identify the trends in the items that our customers want and be able to predict the timeframes for those purchases.

“This requires arming our sales team with the information they need when speaking with their customers. That data has to be correct and up-to-date. Analytics has given us a big part of the picture and AI can take it further. For example, we’ve piloted the use of digital assistants where the sales person can ask for the latest information either before they meet the customer or while in the meeting. It helps make the sales person more productive and improve customer satisfaction.”

The digital assistant pilot shows how AI can begin to spread organically within an organization. “I’ve spoken with our finance team about the work we’re doing with digital assistants. They were excited that they could improve a process of, say, paying an invoice where they could ask a couple of questions without switching applications or interrupting their workflow.”

Supply Chain Use Cases 

Three emerging use cases for AI in the supply chain are pervasive visibility, which is the ability to see exactly what’s happening across the entire chain; proactive replenishment, the ability to automatically order and receive new parts and material exactly when they are required; and predictive maintenance.

The steel company’s manager identifies pervasive visibility as a key benefit. “If you look at how our company operates, we source our raw materials from many different places around the globe,” he says. “The geopolitical environment can have a major effect on our business. How are prices and quality going up and down from our different global suppliers? How easy is it for us to switch supply and when should we do it?

“Weather patterns are a great example of where AI can help. Shipping our raw materials to our plants can be hugely affected by weather patterns. We need to work out quickly how the weather is going to affect us and whether we need to change our point of supply to ensure business continuity and that we’re not introducing delays to the final delivery to customers.”

To achieve these gains in the supply chain, AI has the potential to bring together both structured and unstructured data from a wide range of sources including IoT devices, plant operations, and external partners to identify patterns linking factors such as demand, location, socioeconomic, weather, and political status. It can form a basis for a new level of supply chain optimization spanning raw materials, logistics, inventory control, and supplier performance and helps anticipate and react to market change

Opportunities in the Plant 

All manufacturers have a focus on optimizing production and many have invested in technologies to increase productivity and efficiency, boost quality, and control costs. As Manufacturing 4.0 becomes more established, AI offers improvements in a wide range of applications. For the steel company, two of the most important are quality and maintenance.

Delivering the highest quality products to customers has been a hallmark for the company but all manufacturers are faced with the challenge of maintaining quality standards in the face of tightening time-to-market deadlines.

Companies are beginning to investigate AI to deliver predictive quality capabilities to spot quality issues and predict potential problems in time to prevent them. It enables production faults such as equipment anomalies and recipe deviations to be quickly identified and addressed.

The steel company introduced an application to monitor steel as it passes through rollers. It checks that the steel remains central to the roller and immediately alerts where there is a deviation. “As steel passes through the roller, it can move to the left or right, up or down, causing a break-out that can be very bad,” the manager says. “The application reads data from sensors in the rollers and advanced analytics to predict where a break-out might occur. This allows us to slow down the rollers before it happens.

“The monitoring application has given us many data points: Why was there an issue? What’s caused a problem? Currently, the application makes us reactive as we can slow the system down to avoid the problem. I see AI letting us move from reactive to proactive. If our reactive is to slow it down, our proactive is to fix it before it becomes a problem.”

If AI is still nascent in manufacturing today,some surveys indicate that it could become mainstream in under 24 months.

Predictive Maintenance 

Recent research estimates that unplanned downtime costs manufacturers an estimated $50 billion annually, and that asset failure is the cause of 42% of this unplanned downtime9. In addition, Deloitte suggests that poor maintenance strategies can reduce a plant’s production capacity by as much as 20%10. Predictive maintenance has become one of the hottest areas within manufacturing.

Predictive maintenance uses sensors to report equipment conditions on an on-going basis. This data can be analyzed to draw conclusions on a machine’s condition and spot the trends and anomalies that make predictive maintenance possible. Beyond reducing unplanned downtime, predictive maintenance can minimize the need for planned or scheduled maintenance as well as offering the potential to automatically trigger a maintenance visit, if a part can be repaired, or a replenishment schedule, when it has to be replaced.

The result is that equipment can remain operational, at optimal levels, for longer periods of time. “For the steel company, downtime can cost us tens of thousands of dollars per hours,” the manager says. “Our monitoring product gives us the data to be effective at predictive maintenance. Combining analytics and AI will allow us to say ‘here’s what we’ve got. Here are all the pieces. Here are all the components. Now what’s next?’”

The Time to Start is Now 

Some manufacturers, especially auto makes such as BMW and Ford, have been quick to adopt AI but most other manufacturers have taken a more cautious approach. As with most technology innovations, companies have preferred to be a fast follower rather than an early adopter. These companies want to see successful AI use cases before implementing AI solutions themselves.

Harvard Business Review has sounded an alarm about this approach, saying: “By the time a late adopter has done all the necessary preparation, earlier adopters will have taken considerable market share; they’ll be able to operate at substantially lower costs with better performance. In short, the winners may take all and late adopters may never catch up.”11

However, implementing AI is not a simple process and solid foundations have to be in place. The Manufacturing Leadership Council’s Factories of the Future survey provides a number of excellent illustrations of this. For example, the survey found that maintenance was seen as the second most immediate application of AI. Yet, today, only 7% of respondents said that their maintenance process was “extensively” digitized.

Without digitization, it will be virtually impossible to properly apply AI to a process. This also places the focus on data quality. “With our monitoring application, it’s all about getting the correct data in,” the manager says. “That’s the key. If your original data isn’t good and you’re basing your conclusions on this, you’re going to be in trouble.”
The Council’s survey showed that manufacturers are looking to digitize rapidly, and companies shouldn’t delay exploring the benefits of AI until they have totally digitized their processes. Instead, an evolutionary approach based on building AI into the systems you already have in place will help reap the rewards that AI undoubtedly offers.

“I think AI today is about small wins,” the steel company’s manager concludes. “You need to show how it can be effectively deployed, how it will benefit us and how it can be extended throughout our organization.” M

1) https://www.opentext.com/campaigns/manufacturing-ai-iot/intelligent-connected-businesses
2) https://www.statista.com/statistics/471264/iot-number-of-connected-devices-worldwide/
3) https://www.opentext.com/campaigns/ai-iot
4) https://www.forbes.com/sites/insights-intelai/2018/07/17/how-ai-builds-a-better-manufacturing-process/#627c4de71e84
5) https://www.manufacturingleadershipcouncil.com/2019/02/08/turbulence-ahead/
6) https://www.pwc.com/us/en/services/consulting/library/artificial-intelligence-predictions.html
7) https://www.pwc.com/us/en/services/consulting/library/artificial-intelligence-predictions-2019.html)
8) https://partners.wsj.com/emerson/unlocking-performance/how-manufacturers-can-achieve-top-quartile-performance/
9) https://www2.deloitte.com/content/dam/Deloitte/us/Documents/process-and-operations/us-cons-predictive-maintenance.pdf
10) https://hbr.org/2018/12/why-companies-that-wait-to-adopt-ai-may-never-catch-up