While few would argue over AI’s potential to change the game for supply chains, many businesses find themselves at a crossroads, uncertain about when, where and how to weave AI into their complex operations.
Faced with endless possibilities, what are some of the steps companies can take to harness the power of AI in their supply chains?
Setting parameters for AI efficiency
AI has computational power far beyond the cognitive capacity of humans and it’s easy to understand how companies might be tempted to throw all the technology at all the problems. But, to achieve a grand vision for AI in the supply chain, it’s crucial to first build a solid foundation.
Parameters are key to the machine learning algorithms behind AI. Determining the right parameters requires understanding the specific goals and characteristics of the problem the AI is trying to solve and places more accountability on the people that develop and deploy the AI models.
Rather than trying to solve for every problem, this approach helps companies focus on solving the most important ones efficiently and effectively. For example, one Kinaxis customer has over 25,000 SKUs and global inventory in the billions. They focused on parameters such as cycle days, lot sizes and number of production plants to help them determine the right mix of regional distribution centers for a particular product line.
Spending the time to define parameters not only enhances efficiency but also avoids trying to boil the ocean. It allows companies to concentrate on the most critical issues, providing a practical and effective approach to AI integration.
Overcoming algorithm aversion through explainability
Studies by experts from Wharton and the University of Chicago have shown that humans are inherently reluctant or sceptical about trusting and adopting decisions made by AI algorithms, often preferring human-made decisions - even in the presence of evidence that the algorithm performs better. This is known as algorithm aversion. But, to achieve a higher level of automation in the supply chain, we need humans to trust the math.
In this context, the concept of explainability becomes a crucial factor in creating a culture conducive to AI adoption, making it clear and understandable why – and how – AI systems make the decisions they do.
At Kinaxis, explainability is a fundamental pillar of our approach to AI – for example, our advanced demand sensing and forecasting solution, Demand.AI, has been carefully designed to improve user adoption and acceptance of ML-generated results with easy-to-interpret visualizations that explain which data features affect demand predictions.
By showing them the "why" behind an AI-generated prediction or recommendation, planners are more likely to trust and embrace the outcomes.
As well as driving more widespread adoption, this allows for increased confidence in decision-making and a healthy balance between human expertise and AI capabilities.
Get ready for GenAI
While the Gartner Hype Cycle for Artificial Intelligence predicts that generative AI will become a full-blown mainstream technology in non-supply chain applications in two to five years, because supply chain models are so complex and specific to each company, the expected arrival into the mainstream is anticipated to be 10 years out.1
While some companies are already identifying use cases and piloting projects such as real-time supplier risk monitoring, this provides companies with a window of opportunity to ready themselves for when generative AI comes of age in the supply chain.
One way to do that is by arming your workforce with new data literacy and governance skills to structure, understand, interpret and work with the large datasets that generative AI relies on.
In the words of ChatGPT itself, “The future supply chain workforce must possess a multifaceted skill set, combining technical proficiency, ethical awareness, adaptability, and strategic acumen to navigate the evolving landscape shaped by generative AI.”
At the same time, the critical role of humans in bringing the 3Cs – context, collaboration and conscience – remains unchanged. AI can’t derive meaning from context, it can’t bounce ideas off others to come up with creative solutions and it can’t tell right from wrong. We’re always going to need those uniquely human qualities to validate and operationalize AI outputs to drive positive outcomes.
Data is another critical success factor. We’ve all heard the saying ‘garbage in, garbage out’ and nowhere is this more applicable than when it comes to generative AI. It doesn’t need to be perfect, but it does need to be decent.
Companies should ask themselves what data they’re capturing and how they’re capturing it. How is the data structured? What is the context of the data? And how do they protect it? If widespread applications of generative AI in supply chain won’t be seen for another decade, one thing you can get started on right now is getting your data GenAI-ready.
AI in action
While harnessing AI in your supply chain may seem like a somewhat daunting prospect, the juice is worth the squeeze.
Take the case of a major CPG manufacturer that was experiencing lengthy consensus demand planning cycles, high levels of bias and poor forecast accuracy. Thanks to our Demand.AI solution, the company achieved a >100% improvement in forecast accuracy and reduced forecast bias by 7%, leading to increased trust and analytical rigor in the forecasting process and a ~50% reduction in manual planning activities.
Now that’s what we call harnessing the power of AI!
For more insights and practical strategies to harness the power of AI in your supply chain, tune in to our Big Ideas in Supply Chain podcast: AI Unleashed: Revolutionizing Supply Chain Innovation.
1. Exploring the Power of ChatGPT: An Opportunity for Supply Chain Transformation (gartner.com)
Leave a Reply