- Supply chain and manufacturing are two functions most likely to see cost savings as a result of using artificial intelligence (AI), according to a recent survey by McKinsey and Company that asked hundreds of respondents about the benefits their companies were seeing from AI and its impact on organizations.
- The survey found 64% of respondents saw cost reduction in the manufacturing space while 61% saw a reduction in supply chain planning costs.
- The report suggests the supply chain savings are the result of better spend analytics and better network optimization. The manufacturing savings were are the result of improving “yield, energy, and throughput,” the report noted.
Manufacturing and supply chain management are complicated operations, which often leads to waste, according to Freightflows CEO Matt Morgan.
“Any time that there’s complexity, machine learning and AI can add significant benefit,” Morgan said in an interview with Supply Chain Dive.
The MAPI Foundation, a group that advocates for the manufacturing sector, says AI can optimize various stages of the supply chain from warehouse management to supplier relationships, according to a recent report.
One example is Procter & Gamble’s use of AI and Internet of Things (IoT) technology to automate warehouses and distribution centers. P&G was able to automate delivery of about 7,000 SKUs and cut supply chain costs by about $1 billion annually, the MAPI Foundation noted in its report.
Noha Tohamy, a distinguished vice president analyst at Gartner focused on supply chain analytics, said the findings of the McKinsey survey were in line with her experience.
The two main benefits of AI adoption are increased efficiency (doing more with less or the same capacity) and decreased cost. The cost reduction is not typically the result of decreased headcount but rather improved forecasting. AI can provide better answers more quickly and “can improve my market share or my ability to make products available,” Tohamy explained.
The use of AI and IoT in the manufacturing setting is often referred to as “Industry 4.0.” Predictive maintenance is one area that could lead to efficiency gains, according to proponents of Industry 4.0. The ability for sensors to monitor machines and pass this data along to algorithms that determine if there’s a problem can decrease downtime of equipment and thus increase output for any given factory. Improved uptime for machinery was the most common goal of predictive maintenance, according to to a survey from PwC.
But Morgan cautions that AI and machine learning are simply tools to reach a goal, not a goal themselves.
“You need to start with a hypothesis, you start with a goal,” Morgan said. “And then you need to use the data to go find out how to improve that metric in the business.” Modern technology has made it economical to apply machine learning techniques at a larger scale, but that doesn’t mean they’re the solution to every problem, he said.
Once the hypothesis is in place, data will be required to test it.
“It doesn’t have to be perfect data and it doesn’t have to be massive and massive amounts of data,” Tohamy said. “But there’s enough data for us to start training the AI model or the machine learning model.”
Deploying a modern, industrial AI ecosystem is more than collecting some numbers and building a model. This environment requires analytics technology, big data technology, cloud infrastructure, domain knowledge and evidence, according to a 2018 paper in Manufacturing Letters by four researchers at the University of Cincinnati. This combined environment includes technology, but also the industry-specific knowledge needed to make AI a realistic tool.
As companies get this mix right, more will want to follow in their footsteps and the market for industrial AI is expected to expand 52% between 2017 and 2024, according to Research and Markets.
Just a few years ago, the examples of companies using AI and machine learning were few and far between, Tohamy said, adding that the outlook for AI in the supply chain space is “extremely positive” but said there was still work to be done to scale the technology across larger networks.
One of the challenges companies face is “scaling up these use cases and those examples,” Tohamy said. “So, it’s one thing to use AI for one specific product or one specific region, but how can I deploy that across an entire global, complex supply chain. These are the areas where I need more proof points.”
This story was first published in our weekly newsletter, Supply Chain Dive: Operations. Sign up here.