Letting AI call the shots
From the February 2023 print edition
The runaway popularity of ChatbotGPT, OpenAI’s portal that was launched in November, has brought the power of AI to millions of online users, prompting a scramble by Google and other major tech players to gain a foothold in AI.
This heightened interest coincides with an urgent need for faster and better-informed supply chain decisions.
AI is already providing guidance about supply chain vulnerabilities. In a joint project, the Association for Supply Chain Management (ASCM) and management consulting firm KPMG have developed an AI-powered index to quantify the stability of US supply chains.
“Our supply chain stability index, which is the first of its kind, was built in collaboration with KPMG using AI and machine learning to measure the stability of the supply chain,” says Douglas Kent, executive vice-president of corporate and strategic alliances, ASCM. “We used about 15-years-worth of data for that and 30 different underlying variables. And what it clearly indicates is that we’re twice as fragile as we were pre-pandemic. That’s a massive jump in a relatively short period of time.”
The report was by no means restricted to the most obvious conclusions. “Traditionally, the contribution to instability comes from big changes in market demand, or big changes in available supply,” says Kent.
“Those are there, but perhaps the biggest shock in the release of the report was that fulfillments – being able to fill orders – was causing the highest degree of instability.”
From data to insight
The big push for supply chain operators is to gain visibility of a much larger range of potentially destabilizing factors. “Everything that companies are asking us for leads to gaining visibility of these factors earlier so that they can react sooner,” says Sherief Ibrahim, GM, business applications, Microsoft Canada.
“We’ve done a lot of work with Kraft-Heinz, so I’ll use that as an example. They had to re-evaluate the mix between eating out and eating in and what that means for packaging, and for distribution channels. So how do they take the insight that they’re getting from all these different systems, and from social sites and other sources, and actually make a decision about how they might change?”
Long, multi-tier supply chains have a particular need for expanding visibility. “We spoke with a CEO of an industrial manufacturing company earlier this week,” says Patrick Van Hull, senior director of industry solutions, o9 Solutions, AI-powered supply chain planning solutions, “and that company is so far upstream in the supply chain, and there are so many tiers between it and the end user, that no data is reliable. So forecasting demand tends to be guesswork.”
Progressing from data to actionable insights, however, is a formidable undertaking calling for a disciplined, long-term approach. “Companies are sitting on reams and reams of data,” says Ibrahim.
“What they need is a data strategy, and that can take some time.”
The key is taking a structured approach.
“The data we’re looking at is extremely heterogeneous,” says Dr. Alexander Wong, University of Waterloo engineering professor, Canada research chair in the area of artificial intelligence, and a founding member of the Waterloo Artificial Intelligence Institute, “so I usually tell people to do this in steps.”
Wong suggests gaining traction with straightforward projects and then progressing to some
of the more complex ones. “If you rely on tweets and other social media posts to predict supply and demand, that’s much harder to get at,” he says, “because that data is very unstructured, and you don’t know how reliable it is. So unless you’re already very advanced, that’s an area I advise people to not get into yet.”
Microsoft recently launched Supply Chain Platform, an AI-powered platform which helps bring these data sources into a single database and then correlates that information with KPIs and other variables. “In Supply Chain Platform, you can connect your internal and external systems and then apply business logic to that data,” says Ibrahim. “For example, it could make decisions based on whether the delivery is tied to an SLA.”
“The major tech vendors have built up substantial AI teams to be able to help manufacturers and other companies build systems that are quite useful and impactful,” says Wong, “but at the end of the day, it still relies on the actual companies to figure out what their problems are so that people can build proper solutions. So it’s not a solution problem – in many cases it’s figuring out what has the most impact on your company.”
Some initiatives are completely off target. “We’ve seen situations where manufacturers are relying on forecasting when their biggest problem is mitigating issues with their production,” says Wong.
The need to articulate needs in the context of data places new demands on supply chain personnel. “The problem is that when you accelerate adoption of new technologies, oftentimes the human skill sets of the individuals who manage them may not have kept pace, and that’s certainly what we’re seeing here,” says Kent. “There’s no organization that I work with today that doesn’t have or is not looking actively for increasing data science and related capability inside of the organization.”
ASCM, Kent notes, has recently upgraded its training programs to reflect the need for a higher level of data literacy.
Automating decisions with AI is similar to automating factory work in that the best ROI comes from freeing employees of mundane and repetitive tasks. “What the technology ultimately does is help companies reduce the time-consuming tasks in order to free up employees,” says Ibrahim. “Being able to identify that you have low stock in one location, high stock in another, and then making a decision to transfer an order from one to the other sounds fairly rudimentary and basic. But when you have to pull so much disparate data to make that happen, it makes sense to let the AI actually make that recommendation.”
The presence of AI in the workplace causes employees to fear that they will lose their jobs, Kent says, but the thrust behind AI is not replacing workers but making them more productive. “If I use these technologies to remove the non-value redundant activities that I have today in my current role,” says Kent, “then I as an individual should remain employed and find more satisfaction in my work, because the work I’m not doing matters more to the company.”
One example of that is that employees who are free from mundane tasks are more likely
to get involved in sustainability initiatives, which might require, for instance, a strategic shift to alternative fuel sources.
Kent hopes that the upskilling of employees will make the profession more appealing.
“Supply chain tends to get undersold at the entry level,” he says. “But if you’re a warehouse worker, and you’re getting introduced to technologies like robotics and machine learning, where else could you gain knowledge of these technologies at the entry level?”