September 21, 2023 Navigating Generative AI in Manufacturing & Supply Chain

Generative AI presents massive opportunities for manufacturing and supply chain, but leaders should tread with caution.

By Edmund Zagorin, Founder and Chief Strategy Officer, Arkestro, and Alan Rice, Managing Director of Caché Procurement

Generative AI has captivated business leaders in the manufacturing and supply chain sectors. From public company earnings calls to shop floors, the mention of generative AI has witnessed a viral spike in recent months.

In fact, companies like IBM are already pausing hiring for roles that AI can perform, underscoring its potential to reshape manufacturing processes. Generative AI can streamline the drafting of technical documents; assist in identifying the most suitable suppliers based on factors like cost, reliability, past performance and capacity; and even address simple supplier queries and answer frequently asked questions. Needless to say, the technology has huge implications for saving time and improving efficiency.

However, amidst the enthusiasm, some have begun to question whether AI is all that it’s cracked up to be, or if we are currently in an AI hype cycle. In this article, we delve into the challenges that come along with this new technology, and how manufacturing and supply chain leaders can be proactive about AI’s limitations and potential risks.

Fraud + Spearfishing.

ChatGPT has raised alarm among security researchers, who believe that the technology has caused a “wave” of phishing scams. We’ve already seen an uptick in fake purchase orders, for example, which can trick suppliers into disclosing sensitive financial information. Once the supplier is compromised, the hack can spread horizontally along the supply chain, infiltrating potentially thousands of other enterprises (just consider the 2021 SolarWinds attack, which used a single supplier’s compromised systems to attack 18,000 other entities).

Education can be a strong antidote to this. Consider starting a training program to help employees identify what a fake purchase order or phishing scam looks like, and make sure that they understand the potential dangers of these types of security breaches. The technology is evolving quickly, so it’s best to hold regular (every six months or so) sessions with any updated information.

Release of Sensitive IP.

Because AI learns from the information it’s fed, any data entered into generative AI tools can, in theory, be spread to others using the interface. The risk for manufacturers is high – there is a potential to not only expose sensitive data and financial information from their own organization, but their purchasers as well. This can leave the manufacturer open to legal consequences, as well as harm their reputation moving forward.

While many employees have the best intentions, they may not know that everything they input into an AI language model is essentially “public”. Make sure that employees know what information should never be shared with generative AI. If non-public information is input into an AI model, it should be anonymized and encrypted to ensure adequate data protection.

Algorithmic Bias.

AI models can also be used for tasks such as supplier selection. Unfortunately, because AI is trained on past decisions, which may include (unconscious) bias against suppliers of certain backgrounds, these models can be inherently biased. This can lead to unfair outcomes and in certain cases, legal liability. We’ve already seen how this has played out in unfair hiring practices, for example.

For this reason, making the AI’s decisions as transparent as possible will be critical. This could involve sharing with suppliers how the AI is making its decisions. Internally, it may also be wise to form a “tiger team” dedicated specifically to AI, which can focus on detecting and mitigating biases in the AI’s data and outputs.

AI-Induced Supply Disruptions.

As AI becomes more fused with autonomous systems, it will monitor markets and conduct buy cycles for repetitive transactions. This increases the risk of causing a shortage by trying to “beat the market” ahead of a price spike. In short, AI could detect or preempt a run on a specific commodity – including necessary goods like food, fuel and medicine – and trigger an autonomous buying cycle ahead of that run, thereby causing it. Additionally, AI-managed inventory decisions could result in too much or too little inventory, backorders, loss of revenues, increased logistics costs, and loss of customer confidence.

To avoid the “panic buying” scenario, start having conversations with your team now as a preventative measure. Questions you may consider include: what commodity categories are potentially vulnerable to the “panic buying” dynamic? How are other market participants working through this challenge? What kind of human oversight measures do we have in place?

Generative AI, with its transformative capabilities, presents a unique opportunity for manufacturing and supply chain leaders. However, success in harnessing its potential requires careful consideration of the associated risks and challenges. By adopting a manufacturing-specific AI policy, companies can navigate the evolving landscape, embracing the benefits of AI while minimizing potential perils.

Edmund Zagorin is the Founder & Chief Strategy Officer of Arkestro, the leading Predictive Procurement Orchestration (PPO) platform. With a background in network analysis and auction theory, Edmund is a globally recognized thought leader on the emerging role of AI/ML in procurement and supply chain.

Alan Rice is the Managing Director of Caché Procurement, an organization that provides procurement advisory services, training, strategic sourcing support, and targeted recruiting to clients who are looking for a tailored cost-effective approach to managing their spend.

Subscribe to Industry Today