Developers of the technology say direct sourcing has been shortchanged by most of the spend-analytics applications in the marketplace today. Enter LevaData Inc., a three-year-old company that set out to build a cloud-based “cognitive sourcing” platform, based on the precepts of A.I.
Recently the young company got a big infusion of cash, in the form of a $5m Series A investment from Tola Capital. LevaData expects to use the money to further refine its platform.
Chief executive officer Rajesh Kalidindi claims the use of A.I. can help sourcing organizations slash direct materials spend by 10 to 30 percent. His prior experience includes a stint as a sourcing executive at Cisco Systems, Inc.
So what is “cognitive sourcing”? Kalidindi defines it simply as the application of A.I. technologies to that particular piece of the supply chain. The tool, he says “helps in sensing opportunities and risk, predicting outcomes based on that information, and recommending action associated with that.” It also learns from experience.
The technology incorporates machine learning, in particular “deep learning,” an approach that extends beyond the use of task-specific algorithms. It also draws on natural language processing, which can help sourcing professionals understand verbal feedback from suppliers, and respond in kind.
Kalidindi says LevaData is out to help companies make sense of the massive amount of data that is inundating global supply chains today, especially in the area of sourcing and supplier management. Sourcing teams have neither the time nor ability to sift through all that information without the help of modern-day technology tools. Their focus needs to be on the critical sourcing decisions that will impact the organization’s bottom line.
“A.I.” can itself be a difficult term to pin down. LevaData works from a generic definition, says Srini Kumar, vice president of product management and data science. It views the concept as a means for computers to mimic the cognitive functions of the human being, as well as the brain’s ability to learn from experience. That suggests a technology ranging beyond the “brute-force” applications of some sophisticated machines, such as those that can beat humans at games of chess or Go.
The early days of A.I.’s application to the business world saw the deployment of “expert systems,” which essentially mirrored the abilities of humans to make decisions based on years of hands-on experience. The term is heard less today, having been supplanted by “machine learning.” Again, the shift suggests that today’s A.I. programs are more than a repository for raw data.
“Because of advances in computing, you’re able to feed huge amounts of data instead of symbols, and it’s literally learning from the data,” says Kumar. “You don’t have to reprogram it to get better for certain tasks.” For example, an A.I.-driven application should be able to improve its ability to predict prices based on its knowledge of past pricing decisions, and their impact on sales.
Cognitive sourcing can be deployed throughout the procurement process, Kumar says. It can perform predictive costing, select suppliers and even negotiate with them. And it can advise humans — who, obviously, are still overseeing the process — as to any important details that they might be missing.
Kalidindi says the platform can be employed to support sourcing for new-product introductions — a traditionally thorny task, given the targeted item’s lack of history. It can also help to assess the risks associated with such elements as end of life, and financial requirements for supporting the new product.
There are, of course, no fixed answers to any given sourcing scenario. Outcomes will always be impossible to predict with complete accuracy. But A.I. is beginning to allow sourcing professionals to better assess probabilities, based on a review of multiple “what-if” scenarios. An 80-percent probability score, for example, will prompt the system to recommend the applicable decision.
Scenarios will differ according to how aggressive the buyer wants to be with potential suppliers. If a negotiation runs out without a satisfactory agreement, that’s a sign that the buyer needs to adopt a more practical stance. Despite the rigors of A.I. science, there’s no strict formula that guarantees consistent success.
“We are not making a religion out of A.I.,” says Kumar. “We’re using it pragmatically.”
Humans retain the ability to override any decision dictated by the system. But even those actions get incorporated into the mix, so that the machine is likely to do better next time.
For all of the hype surrounding A.I., Kalidindi says it’s more a provider of “augmented” intelligence at this stage of development. Over time, he believes the system will evolve from supervised to unsupervised learning.
One rule of thumb states that if a procurement professional is not regularly speaking with other humans for a given process, “then that job function is at risk of being automated at some level,” says Richard Barnett, LevaData’s senior vice president of marketing and customer success. Examples include invoice reconciliation and purchase-order matching based on simple rules.
Still, sourcing will always remain to some extent a game of relationships, for which people remain eminently qualified. “Humans,” says Kalidindi, “are not going to be cut out of the loop.”