At the ToolsGroup North American User Forum earlier this fall, the company’s CEO talked about how ToolsGroup is incorporating machine-learning technology into its demand planning solution. Machine learning, which is a branch of artificial intelligence (AI), uses specially-designed algorithms to generate predictions based on entered data. ToolsGroup is using AI to improve its demand forecasting capabilities. For example, the company is looking to see if it can create variables based on new sources of data, such as social media, to improve demand forecasts.
Historically, companies bought demand management solutions developed by smart humans, and then implemented the algorithms that came prepackaged in the solution. For a while, these solutions were periodically tuned by humans to make sure forecast accuracy did not degrade. But newer demand management solutions have “self-healing” capabilities — i.e., they can look at historical forecasts, and the accuracy of those forecasts, and then suggest using a different prepackaged algorithm in particular situations.
I recall the first time I heard about self-learning software programs. I had a hard time wrapping my head around the concept. But these programs exist, and self-healing demand management solutions are a good example.
But AI solutions that build complex supply chain models autonomously? A technology that replaces the need for smart analysts? Wow!
I asked ToolsGroup if I could talk to a customer using AI in this manner, but at this point, the company is not ready to provide references. But one of my long term industry contacts, who was also at the forum, told me that Dell is using AI technology, provided by a different vendor, in its global command centers. So I reached out to and spoke with Owen Panko, the Director of Program and Project Management at Dell’s Global Command Centers. Dell has five global command centers, analogous to NASA’s mission control center, staffed with about 200 people globally.
These control towers use business process management (BPM), AI, and analytics as core tools to support their service delivery supply chain. These command towers provide visibility and process flows to parts, people, call center activity, and their technology resolution experts. For large corporate customers, their standard service level agreements include same business day, two hour, and four hour response times.
They have a tool they’ve built on top of a BPM solution that provides business process rules at the transaction level. A particular type of service request, for example, might need to be routed to a technician with a certain skill set. The BPM level of their solution also provides alerts. For example, if a dispatch order includes both a motherboard and a processor, an alert is raised to take another look at that order because including both items should really not be necessary.
On top of the BPM level runs a complex event processing (CEP) engine. CEP is also a form of AI. The CEP layer looks at the thousands of alerts generated at the transaction level and looks for larger patterns. For example, Dell expects to receive acknowledgements from 3PL partners when they receive a service request. If Dell is getting a myriad of non-acknowledgements, the CEP engine knows to look to see if the non-acknowledgements are all coming from the same 3PL. If so, it makes sense to contact that 3PL and have them correct the problem. Or the CEP might detect that call volumes have dropped in a certain region. One of their managers is flagged, and with a couple of phone calls he might be able to provide an explanation (perhaps there is a holiday in that region, for example).
Now, as far as I was concerned, the critical question was this: Did the CEP engine find these patterns itself, or did analysts program it to find the connections? Owen says they have an analysis team that includes statistical experts. These experts look at historical trends, deviations from expected service levels, and they look for significant patterns. They can then program the CEP to detect and respond to these deviations in real time. So, in short, Dell’s AI solution is not a self-learning entity capable of building or augmenting complex supply chain models autonomously.
However, that does not mean that what Dell is doing is not cool; it is very cool. The company’s command centers, and the technology that supports them, provides a competitive advantage. Thousands of customers and prospective customers tour these command centers every year. They see what Dell is capable of, and many decide that Dell is the right partner for them.
We’ve written about robots replacing people in warehouses. But apparently, supply chain analyst jobs are safe for a little while longer.