CEO of Edge Platforms, EdgeVerve Systems Limited.
Can artificial intelligence (AI) make supply chains more agile? The short answer is yes. For a fuller explanation, let’s first consider where supply chains stand today.
With deep roots in supply chain, I’ve been through many cycles of innovation, deployment and inertia. These 25-plus years of entrepreneurial, management consultant, and operational and IT experience have helped me to assess the transformational potential of AI/ML within a deployable reality.
Supply Chain Impediments
Take the vision of a centralized, cross-functional, supply chain “brain.” While compelling, it remains largely aspirational. Two years ago, for instance, a survey of global supply chain leaders found that close to three-quarters of respondents relied on spreadsheets for supply chain planning. More than half still used an application that is approaching end-of-life (EOL).
Though many companies are modernizing their supply chain IT, impediments and technical debt remain at large. This manifests in redundant or out-of-date enterprise resource planning tools, like the one alluded to above, and also disconnected supply chain processes and point solutions that were once quick fixes and now need more work.
It sometimes takes a crisis to bring technical debt to the surface: You receive the EOL notice from your vendor. Or a legacy software application needs work, but no one uses that code anymore. Or your developers are too busy with workarounds to deploy anything new. Patchwork complexity can make supply chains brittle and difficult to refresh.
How AI Can Boost Agility
No technology is a panacea. But here’s how AI/ML could address some of these concerns and help create an integrated supply chain platform that delivers flexible and agile outcomes:
1. Identify and remediate debt. We can use AI to identify burdensome technical debt. Tools such as generative AI can help further by extracting and reading outdated code, as well as by fixing and refactoring it. Reducing debt can free up resources you may need to undertake a supply chain platform upgrade.
2. Extend visibility. AI can improve visibility and correct for information deficits. Adding data is one way, but not the only way, to improve AI. That said, AI/ML opens the door to higher-volume and more diverse data from suppliers and consumers. These tools can also receive data on macro trends, thus increasing awareness and reducing the potential for external shocks.
3. Harmonize data. A common problem on the granular level is the lack of standard classifications. Supply chains may encompass millions of SKUs. Tracking their movement is one challenge, but inconsistent identifiers compound complexity. You could have visibility but no intelligence—and up until now, no quick way out. But AI/ML changes the game, enabling you to harmonize master data in minutes.
4. Process data. Data processing can occur at various points along the supply chain. At the edge, you can use AI/ML to classify data, discard null sets or identify anomalies. As AI-enabled processes morph into a kind of central brain, you’re able to analyze a myriad set of variables. Sorted and ranked outputs provide a shared reference point for next actions.
5. Deliver actionable insights. The proof of agility lies in cross-functional execution. A marketing team knows in near-real time what consumers are purchasing and adjusts their campaigns as a result. Similar insights could drive suppliers and distributors to optimize inventories under their control. Operational benefits extend to other third parties.
Best Practices
My experience includes implementation of AI across multiple industries, not just at the proof-of-concept level, but also at production scale. For example, I worked with a sports fashion retailer that created a connected network of suppliers, manufacturers, wholesalers, distributors and retailers to reduce inventory holding and respond quickly to changes in demand. By linking multiple roles to a cloud-based, AI-powered platform, they delivered benefits to each partner in the value chain and enabled the retailer to fulfill demand from anywhere and precisely target micro-markets and Zip codes.
From my perspective, successful AI initiatives like this have two things in common.
1. Buy-in from all stakeholders: Make sure to include external counterparts or internal leaders from IT, marketing, risk management, logistics, etc.
2. Recognition of the costs: Models can become complex and AI engines have their own upgrade cycles. They require computational resources and skilled human managers who provide ongoing attention to data quality and bias.
Challenges
I’ve mentioned a few of AI’s operational impediments in passing, such as technical debt and inconsistent terms. Without addressing these foundational issues, you will be limiting your return on any investment in AI.
But AI has its own challenges:
1. Expertise: Supply chain leaders need some in-house knowledge—especially around generative AI. Deployment scenarios could involve foundation models trained on massive data sets, model aggregators and open-source models. For those lacking expertise and resources, independent model hubs are a good option. These may provide access to ML operations tools and allow you to fine-tune a foundation model with proprietary data.
2. Security and privacy: Without sound governance and mitigation techniques, some implementations of AI can put internal and partner data at risk. Restricting the use of AI by employees, at least until you have set up guardrails, is one option. Supply chain IT leaders will need tight security around their use of AI to gain the trust of—and data from—external partners.
3. Other concerns: The unprecedented power of AI has some calling it an existential threat to humanity. On an enterprise level, AI does create additional liabilities. Trained on historical data, for instance, it can generate results with internal bias. If not corrected, the bias could be self-perpetuating. Automated loops on purchase decisions—another unintended result—might lead to disruptions of supplies. Built-in checks, including reinforcement learning from human feedback, could preempt or limit such outcomes.
Realizing A Vision
The quest for end-to-end supply chain visibility and health has a long history. Supported by powerful AI/ML algorithms, this journey is looking less quixotic and more within reach.
Implementing an AI-enabled supply chain is far from a set-it-and-forget-it exercise. Like any meaningful IT project, it requires focus and attention. Active buy-in and cost recognition are critical. But the potential benefits are real. The actionable insights that follow can drive heightened responsiveness, flexibility and efficiency.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
Read the full article here
