When it comes to Artificial Intelligence, many companies are still in the learning phase. They’re trying to understand the technology landscape and what’s possible, but it’s a bit challenging because everything is evolving so fast. At the moment, the buzz and hype are centered on Generative AI, thanks in large part to the November 2022 public introduction of ChatGPT. But the AI field is actually much older and broader than that. So, why is AI getting so much attention today? What benefits does it promise? What defines a good AI strategy? And what are some lessons learned for successful deployment of AI in your supply chain? Those are the key questions I discussed with John Lash, Group Vice President, Product Strategy at e2open, on a recent episode of Talking Logistics.
Why the Hype?
I began by asking John why there is so much hype now, and what the benefits really are.
John notes that there is a lot of interest in AI currently, but he has been working with it for 15 years, starting with demand sensing. As head of product strategy at e2open he talks regularly with clients about their strategies and use of AI. “Regardless of where they are in their journey, there is a universal belief that AI has enormous potential to be transformative, but maybe not in the way most people think, given the hype.”
John comments that AI has been woven into the fabric of many systems we rely on today. These AI applications run quietly in the background. What’s different with generative AI like ChatGPT is that it is conversational. This allows people to easily interact with it in many ways. That’s where the hype comes from.
“The real benefit of generative AI comes below the surface in unlocking unstructured data for use by traditional AI. It’s like an iceberg where generative AI is the 10% above the surface, but the real benefits come from 90% below the surface which is traditional AI. The corollary to that is you need good traditional AI to realize the full benefits.”
Traditional vs. Generative AI
John explains that there are four classes of AI: Supervised, Unsupervised, Reinforcement and Generative. The first three make up traditional AI. They work in the background to create new value beyond what people can do. Watch the short clip below where John defines Supervised and Unsupervised AI:
Supervised AI, which is the most widely used, learns by finding patterns across disparate data sets to accurately predict outcomes. Unsupervised AI discovers hidden clusters of data without any training or guidance which may not be obvious to people. Reinforcement AI explores potential outcomes on a trial-and-error basis to discover which actions produce the best outcomes.
Generative AI uses large language models to interpret masses of unstructured data to generate new content with similar characteristics. This often features human-like interaction capabilities that made ChatGPT famous.
“These are four different AI tools and they each do different things that have their own uses and value. It’s a matter of using the right tool for the right job,” John says.
Given the four types of AI, I asked John what defines a good AI strategy. John states that you should start by having all four in your toolkit. But he says there are four prerequisites for a successful strategy.
The first is data. AI needs large amounts of data that is relevant to the business decisions at hand, as John discusses in the short clip below:
The second prerequisite is privacy and security to make sure sensitive data does not get into the public domain. The third is transparency, because you want to be able to trust the results are valid. “Without trust, people override the AI and go back to doing things the old way.” The fourth is closed-loop orchestration. Once you make a decision, how do you put that into action? That’s the crux of decision automation.
John notes of these four, privacy and transparency are table stakes for vendors. It is in handling data and orchestration that AI products separate themselves. The data should come from throughout your supply chain and ecosystem partners. “That’s where a supply chain operating network’s ability to supply good data to your AI is so helpful.” Closed-loop orchestration is important because, “Without the ability to act, AI decisions are just wishful thinking.” Again, the benefit of a network is the ability to put decisions into action.
Finally, generative AI is so useful because it can make unstructured data, which accounts for about 90% of all data, available to traditional AI to make better decisions. “Companies that can successfully unlock this data will perform better than companies that can’t.”
Applying AI for Better Decisions
The number one goal for business executives is making better decisions. How does AI facilitate that? And what are the lessons learned and advice from the early adopters of AI? John shared some great insights on those questions and more, including a discussion on “corner cases,” so I encourage you to watch the full episode for all the details. Then keep the conversation going by posting a comment and sharing your perspective on this topic!