Robert Kugel's Analyst Perspectives

Resilient Supply Chains and AI to Manage Costs

Written by Robert Kugel | May 23, 2024 10:00:00 AM

Rebalancing supply chains to improve resiliency has been a focus of enterprises with even moderately complex and long supply chains for the past four years. One aspect of this rebalancing is that it almost always involves higher costs. Volume discounts and bargaining power are reduced when more suppliers are used, or an alternate supplier may have higher factor costs and therefore must charge more. Logistics costs might increase, and when an enterprise moves from just-in-time to just-in-case for some or all production inventories, it will incur higher expenses. This includes higher working capital costs, especially in a world where real interest rates are higher. Economics dictates that doing business is always a matter of managing trade-offs — in this case, changing from a singular focus on supply chain costs to optimizing cost against adaptability and therefore resiliency. However, the history of human endeavor and economics shows that a third dimension — technology — also plays a role in handling trade-offs, enabling enterprises to achieve an optimal trade-off at a higher level of utility. In this case, allowing enterprises to minimize the relative costs of successfully adapting to change. The upshot is that enterprises should use technology to achieve competitive advantage.

Some years ago, I started commenting on supply chain management in a new era of trade. That was based on an observation that the world had been shifting from a half-century trend of increasing trade liberalization to a different world order that was increasingly less so. In this environment, enterprises would be dealing with a more complex set of trade regulations, tariffs and taxes. Consequently, for many businesses, the lowest-cost supply chain paradigm that was possible in a liberal trade environment was no longer an ideal strategy. This became obvious during and after the pandemic when “supply chain disruptions” became a news item. Measured by the volume of Google searches, during 2019 the term averaged an index value of 65.3 (against a maximum of 100), which rose to 70 in 2021 and 81 for all of 2023. So far in 2024, it has registered at 88.

A central requirement for supply chain resiliency is an ongoing process for a high-level review of the strategic elements that will impact this part of an enterprise’s business model. Without a clear strategy, what passes for agility is just flailing. The strategic decisions set the conditions under which cost optimization efforts can be made, but that is not my focus here. Once management sets a strategy, though, the organization must have the tactical ability to swiftly make well-coordinated and informed decisions in day-to-day operations to get the right balance between achieving business objectives and minimizing total costs.

Resiliency measures the ability to deal with change, especially that with a negative impact. There are two sorts of interconnected components to improve supply chain resiliency. One is having tactical forecasting and planning processes that can chart a future course that maximizes optionality and provides deeper visibility into future challenges and opportunities. And since the purpose of planning is to improve decision-making and outcomes, the second is better supply chain execution: having the ability to adapt faster and more intelligently to changes in the supply and demand environment. Embedded artificial intelligence (AI) in supply chain and S&OP software can enable enterprises to be more resilient and better optimize their decision-making. The adoption of these technologies is just starting, and uptake is likely to be slow. ISG-Ventana Research asserts that by 2026, one-fourth of enterprises will routinely use AI in accelerating execution of sales and operations planning cycles to improve agility and increase focus on more effective decision-making.

Supply chain planning and S&OP are notoriously complex processes where the sheer volume of the minutiae that must be considered is a constant distraction from seeing the big picture. One constant source of complexity is having to harmonize the different perspectives of groups involved in planning and execution. For instance, in many enterprises, the sales organization is focused on meeting revenue targets and creates account plans expressed in revenue potential. Accurately translating those money-based projections into a unit-based set of stock-keeping unit (SKU)-level forecasts is hard enough but is made even more difficult when planning the logistics of getting finished physical goods to customers at monthly, weekly or daily intervals. Add to this the need to figure in the profitability of products and channels, essential for making economic trade-offs. In theory, analytic tools should be able to create useful models that cut through this complexity in ways that human minds cannot. Until now, modeling all of this was impractical from a technical standpoint, but AI using machine learning (ML) and generative AI (GenAI), as well as the ability to employ cloud-based elastic computing resources, are making it technically feasible. That said, what we’re about to learn is to what extent, and under what conditions, technology can provide useful intelligence and how much this can compress decision cycles.

George Box famously observed, “all models are wrong, some are useful.” There are three sets of hurdles that enterprises need to surmount if they are to make effective use of modeling technology in advanced supply chain planning and execution:

  • Dealing with false positives.
  • Creating and using a holistic approach to incorporating technology in their operations.
  • Executing rapid planning and decision cycles.

False positives (or type 1 errors) are a frequent issue in the application of statistical analysis to real-world problems. This is especially true in business or medicine where systems are often not deterministic enough to lend themselves to simple modeling. In practical terms, this means that the value of using technology to provide sufficiently rigorous oversight is undermined when the detection threshold is set low enough to ensure that the system doesn’t miss items, but then overwhelms human operators with a high volume of results that incorrectly indicate there is an issue when it doesn’t exist. Even when the usual palliatives are applied to addressing this issue (such as having a sufficiently large sample size or structuring data to increase its relevance), enterprises will find there are limits to what technology can do. We will be learning what those limits are over the next five years.

Every business problem requires addressing the people, process, data and technology dimensions in a concerted fashion. So, it’s likely that the usual issues that arise when insufficient attention is paid to any one of these will hamper the effectiveness of using technology to improve supply chain operations. The most likely are data sufficiency as well as the change management aspects that are essential to improve performance.  

Similarly, in business, the elements of optimal states are fleeting, so optimizing supply chains requires a constant reassessment of conditions and how best to cope with them. Forecasts are rarely accurate, if only because the conditions that were expected at the time of the forecast usually turn out to be different. Given this reality, what’s important is having the ability to quickly adjust forecasts and plans as events unfold. As conditions change, what distinguishes being agile from the often counterproductive “firefighting” responses is the degree of effort put into realistically considering different scenarios and their outcomes. AI technology, especially as it uses ML to craft and continually refine forecast models, can substantially reduce the time and effort business analysts must now spend on the mechanics of forecasting. With that, they will have additional time to run a more thorough set of contingency plans that can inform decision makers ahead of time with potential outcomes so they can determine how best to respond. The best responses being those that optimally make trade-offs between financial and non-financial objectives. And because less time is spent on the mechanics, enterprises can review and revise plans at a greater frequency. In turn, shorter planning and execution cycles allow organizations to spot negative and positive trends sooner, promoting agility. That’s the potential, but it remains to be seen how well technology will be used in practice.

Business challenges never change but the tools available to conduct business evolve and give those able to take advantage of their capabilities a competitive advantage. Today, we are on the cusp of a potential radical change in sales and operations planning and supply chain execution provided by AI. S&OP and supply chain execution are minutiae-heavy efforts, where predictive AI and generative AI could produce major improvements in performance and outcomes. However, realistically it’s likely that like many technological advances, it will take time for enterprises to be able to fully take advantage of AI-enabled software. Organizational silo-ed thinking will need to be overcome as well as having the right incentives that properly align decisions. In particular, aligning the financial planning and analysis organization with S&OP as well as integrating sales department revenue planning with unit-based product or service planning. Recognizing the difficulty of achieving this goal, ISG-Ventana Research asserts that by 2027, only one-fourth of enterprises with even moderately complex supply chains will closely coordinate their S&OP with FP&A efforts. Those that do will achieve better performance.

To deal with a new environment of trade and commerce, senior executives and functional managers will need to address the related data, process and people dimensions that are a necessary part of any successful change. One the data front, enterprises must address their existing issues related to internal data quality and completeness as well as external data reliability. With the right data to train AI models, the technology could enable planners to shorten plan-to-execute cycles as well as react faster to demand and supply signals.

Leadership teams in enterprise with even moderately complex supply chains must examine their current strategy to determine how these align with revenue and profitability objectives. They especially must focus on how they can use technology to make winning trade-offs in managing their supply chains. And they must focus on the change management challenge of radically shortening forecasting and decision-making cycles to achieve resiliency and agility.

Regards,

Robert Kugel