A key challenge for manufacturers is connecting integrated business planning (IBP) – a longer term plan – to operational planning and execution – what needs to be done in the near term. IBP is a collaborative process involving diverse business functions that is designed to balance demand against supply in a manner that maximizes the goals of that company. This is, at many companies, a month-long process that produces a supply chain plan that spans out over several months – often 24 or 36 months. For the supply chain team, it is the plan for the coming month that is most important. It is the execution of this plan that will delight or disappoint customers. The problem with a plan that spans out over a month is that stuff happens. Orders can be cancelled, rush orders come in, factory machines go down, port congestion occurs, and so forth. In short, a whole host of activities occur that makes it difficult to delight a company’s customers. The supply chain team must grapple with a couple of questions. What part of the plan should be executed? What new things not in the plan need to be done?
This is where an interesting company called Aera Technologies provides a solution. This private company, funded with $170 million in venture capital, is using artificial intelligence to help automate the decision making around these unforeseen problems.
Aera is using data crawlers to crawl across billions of rows of transactional data on a monthly basis. This is combined with data from external sources on weather, logistics lead times, and sustainability performance. These crawlers never stop working. They are designed to have minimal impact on the performance of the underlying source systems. Machine learning algorithms are combined with business domain expertise to help make intelligent just-in-time decisions that help a supply chain operate smoothly.
Aera Technology calls itself a “Decision Intelligence” company. I call them a supplier of supply chain planning software since many of the problems they solve are supply chain problems. Henry Hwong, the vice president of product marketing at Aera, told me: “The supply chain is full of messiness and volatility. We believed if we could nail decision making within the supply chain, everything else would be a little bit easier to handle. So that’s why a lot of our initial customers and initial use cases are supply chain centric. But we’ve always been building out this technology with the thought of being able to do other things around procurement, finance, and revenue management.”
In achieving supply/demand rebalancing, the solution is looking at what kind of tradeoffs should be made to maximize financial goals while minimizing risk and supply chain disruption. When reality differs from the forecasted plan, the solution is programed to make certain decisions in a fully automated manner, some decisions that are either approved or rejected by humans, while some decisions still need to be made solely by humans. The company claims that for one global consumer goods company the solution was making 12,000 recommendations in a month; 74% of those recommendations were auto accepted.
I believe this claim. If it is the same consumer goods company that I think it is, I saw their vice president of supply chain speak at a planning conference. While she did speak to the specific numbers mentioned above, she had high praise for Aera Technology. Some of her planners were rejecting too many of the suggestions coming out of Aera. But the machine generated suggestions were better than the actions the planners took in most cases. For example, a suggestion from the system might be to transfer 96,154 units of a stock keeping unit to a distribution center in Denver to fill a projected backorder. So, for some of the worst offenders, this vice president took away their power to reject the suggestions coming out of Aera.
But the problem is not only that humans do not always make the right decision, humans too often are not making any decision at all! Between tighter labor markets, ecommerce, and increased volatility in supply chains, a growing percentage of exceptions are not addressed. Even if a company does a pareto analysis and addresses 80% of the exceptions, that 20% of issues unaddressed leaves a lot of money on the table.
This is not an all-encompassing form of artificial intelligence like Skynet in the Terminator movies. Part of Aera’s implementation of their platform involves evalutating each particular business problem that they are trying to solve – whether it’s inventory levels or logistics optimization or sustainability – and creating a specific data model to solve that problem. In many cases, the Aera solution runs on top of a planning system from a leading supply chain planning vendor that is already in place. This incumbant solution might come up with the monthly demand plan, for example, but then the Aera technology is adjusting that plan based on what is happening in the market.
The solution is also not all encompassing in the sense that math is not used to solve all the issues that arise. Allocation is a good example of this. If there is not enough inventory to meet the demands of all the customers, which customers does a company prioritize? One business might decide to prioritize Walmart over all other customers. Another company might decide to prioritize the most profitable customers. These are rules-based decisions, based on input from business executives. The fancy math might determine that there is a looming inventory shortage, but not have the authority to make the decision on how to solve that issue.
When I first heard the term the “autonomous supply chain” a few years ago, I thought this was a pipe dream. Aera Technology is showing me how mistaken I was.
