Accelerate supply chain value with a targeted strategy for AI

There is a widely-held belief that massive data migration and a modern supply chain application stack are required before being able to reap the benefits of AI. While certainly beneficial, these are not fixed requirements. To accommodate data readiness and IT system hurdles, companies can incrementally deploy AI to achieve incremental value within months, not years.

AI implementation follows an agile approach that allows capability to be delivered in increments. This means that it can be deployed with partial data, a subset of algorithms, and gradual levels of automation so that value can be realized sooner rather than later.

A supply chain planning organization launched an initiative to sense customer demand with AI, delivering capability in four increments. The first increment produced an algorithm that interpreted market signals such as local events, weather, and social trends. Within four months, planners were using a new forecast that beat current accuracy by 5 percent. Three months later, second and third increments were delivered that accounted for price and promotion elasticity, increasing accuracy by another 2 percent.

For the two years that followed, AI was used to reap inventory benefits in the short-term, while a longer- term initiative to implement a modern supply chain planning system was underway. It was not until this stage where they delivered the fourth and final increment, embedding and automating AI in the new system to deliver a more cumulative outcome.

The supply chain workforce continues to struggle with challenges around labor attrition, activism, and availability. Over 73,000 manufacturing, transportation, and warehouse workers participated in over 80 labor strikes across more than 100 U.S. locations this past year.⁵ This incurred productivity losses that exacerbate longstanding operations issues in employee absenteeism, worker rights, and organizational turnover.

As facilities continue to develop and pilot robotic automation strategies for the long-term, the organizational right-sizing actions being taken today are not enough to keep up with market volatility in the short-term. While AI is expected to enable worker productivity gains through use of generative algorithms,⁶ it can also help avoid productivity losses by anticipating employee behavior and building resilience to labor variation.

A distribution organization experienced over 30 percent in annual turnover of its warehouse associates and truck drivers. As labor issues emerged, predefined retention actions were taken but often late to be effective. This led to perpetual overtime labor cost, recruiting expense, and productivity loss in nearly every case of employee turnover. To avoid labor risk altogether, AI was used to help shift the organization from reactive retention to proactive prevention. The algorithm was trained on what variables influenced employee turnover, including compensation, working conditions, and competitive market behaviors.

Within eight weeks, the AI was able to more accurately anticipate when and why an individual would leave their job. This enabled management to become more effective in retention efforts and target a 40 percent reduction in attrition.

Sources: ⁵ILR Labor Action Tracker, Cornell University, Dec 2023; ⁶Generative AI in supply chain: A path to better returns, KPMG LLP, Oct 2023