Agents offer almost effortless automation, so repetitive and boring tasks get done with the least amount of work and, perhaps, more consistently. Agents are an important evolutionary step in the design of business software, similar to the transition from procedural programming to event-driven programming that accelerated in the late 1980s. That paradigm shift created the business software ecosystem we know today because it made applications more flexible and responsive in replicating task performance.
Adding agents to the considerable body of well-developed business software takes the capabilities of these applications to the next level.
We live in an age of uncertainty, not unpredictability. AI has spawned a revolution by making predictive analytics more accessible and practical. Using machine learning, AI analyzes historical, detailed sales data, market trends and external factors such as economic indicators to create and continuously test and validate forecast models that make it possible to predict future demand more accurately. What was once dependent on the analytical skills of the business or financial analyst can now be left to the software to handle time-consuming manual tasks. This allows analysts to concentrate on using their experience and knowledge of the business to produce more reliable forecasts and respond faster and more effectively when the unexpected happens.
Agentic AI orchestrates the multiple tasks necessary to create, test and modify supply chain planning models. These run the gamut from inventory monitoring, sales order analysis, purchasing, inbound and outbound transportation and distribution center optimization. Agents can compress planning cycles by automating the time-consuming, routine work. Shorter planning cycles are a prerequisite for faster decision and execution cycles.
Greater productivity in planning means enterprises have more time to consider contingencies and how best to react when, as usual, plans and forecasts turn out to be wrong. Our research finds that relatively few enterprises have any sort of ongoing contingency planning beyond a high-level base case with broad upside and downside assumptions. One reason is that, until now, basic forecasting and planning have crowded out time available for understanding the impact of less probable scenarios. Contingency planning in business is useful for the same reason airline pilots use simulators to practice dealing with highly improbable but life-threatening events: developing a muscle memory for dealing successfully with the unlikely. Agents can manage the intricacies of contingency planning, simulating abnormal but not uncommon supply chain events, such as supplier shutdowns, port closures and demand spikes, keeping the underlying data science in the background.
AI-enabled supply chain planning systems have limited utility without accurate, consistent and useful data.
The availability of a rich set of data means that these models can be built around multivariate regression analysis so that when real-world results deviate from projections, planners can isolate the specific factors that deviated enough to provide useful adjustments and do so quickly enough to have a positive impact on performance. For example, if landed costs spike because of an unexpected jump in container shipping rates, or demand for a product is higher because of a social media event, planners can quickly isolate causes and work out the best response.
Agentic supply chain management systems can also shorten decision and execution cycles. These systems operate autonomously to continuously monitor supply chain operations and generate action alerts when scheduled activities must take place or when disruptions are detected. Agents can then recommend or execute actions, orchestrating and coordinating the activities of one or multiple individuals involved. Among many potential use cases, inventory agents can identify and rectify unmet demand issues, deal with shortages by recommending and coordinating inventory allocation decisions, identify issues with bills of material and identify opportunities to rationalize parts and supplies. AI can improve warehouse operations management by continually optimizing labor scheduling and periodically adjusting floor plan layouts to reflect product lifecycles or seasonal shifts. Technology can monitor existing or potential supply disruptions based on events or trends, as well as further enhance transportation plans on shorter cycles.
Over the past decade, the demanding environment for supply chain planning and execution has forced enterprises to take a more strategic approach to managing these processes. The need for resiliency and adaptability in a rapidly changing world economy almost always raises costs unless organizations devise ways of mitigating the impact. AI in all its forms can provide the ability to plan, react and execute with greater precision in shorter timeframes to promote agility. It can also devise new ways to reduce inefficiencies in a highly complex process.
I recommend that organizations with even moderately long or complex supply chains review supply chain planning and execution processes to assess areas where AI and agents can achieve a step-change improvement in performance. For executives, it’s important to recognize that software is a necessary but insufficient component of a supply chain AI strategy. Investments in data governance and management are needed, as well as economic, industry and other external data to ensure model fidelity. Change management will also play an important role in ensuring the technology can be utilized to its fullest. While it’s often said that business conditions have never been more challenging, I disagree. Business challenges are constant—the tools available to deal with them change and provide those that take advantage of their potential with the ability to outperform and outcompete the laggards.
Regards,
Robert Kugel