Brenda Westbrook and Caleb Thomson are senior director analysts with Gartner’s Supply Chain Practice. All opinions are the author's own.
The groundswell of excitement we have seen with recent developments around generative artificial intelligence has created a level of expectation that supply chain leaders will begin to use AI and other advanced technologies to deliver new levels of efficiency and growth for their organizations.
This development may also have shone an unwelcome spotlight into the supply chain planning world, where many have yet to build proficiency in the requisite building blocks to take full advantage of both next level AI solutions coming down the pike, as well as established planning tools that already have AI capabilities embedded within them. The good news is that supply chain planning is ripe for experimentation with a variety of AI and machine learning use cases today, and small steps can add up to big results — if you start now.
Even before the most recent wave of excitement around ChatGPT and generative AI, supply chain leaders ranked data-centric AI and ML as their top digital priority among technologies that they are piloting or deploying, according to respondents in Gartner’s Digital Business Impact on Supply Chain Survey. The survey was conducted from March to April 2022.
The intent to implement is clearly there, but when it comes to where to begin, supply chain leaders face a dual challenge. They must strengthen their existing data management practices to effectively leverage AI and ML technologies, while also identifying the right opportunities to deploy these solutions across their planning processes.
For leaders grappling with these challenges, now is the time to act. Supply chain organizations that do not implement AI and ML supported planning processes will be at a competitive disadvantage, struggling to derive insight into market dynamics that can support related agile decision-making.
How to find an entry point for AI and ML
Don’t overcomplicate your initial use of AI and ML technology. Instead, try to leverage existing technology for planning analytics that may not require further system development or data. This minimizes any financial risk, can provide an initial benefit from automation and builds organizational momentum for further investment and deployment. Systems that are already established within your organization also provide a good learning environment for training planners.
A good entry point to deploy AI and ML is in support of ongoing data-quality validation and management, which can include sales and operational history data cleansing and product relationship analysis and alignment. Beyond data validation and cleansing, existing solutions may provide clear opportunities for planners to build comfort and awareness with decision automation. Try to drive identification of opportunities for greater decision automation use-cases by considering what data may be most valid or helpful, and then validating if it is, or can be, readily available.
Along the way, it’s essential to engage planners as technology end users to review recommended AI and ML outputs, to build confidence in the related techniques and drive adoption of relevant decision augmentation into the decision-making process.
How aligning expectations can improve decision making
Try to avoid thinking of AI and ML in planning as an objective in and of itself, but rather as a tool to improve decision making efficiency in the supply chain. Often the organization’s expectations can be misaligned due to inexperience or lack of industry use case adoption by peer organizations. By clearly socializing the objectives of decision augmentation and automation, defining the objectives, and aligning expectations across business and technology leadership and planners, efforts can become more focused and productive.
An objective could be to improve demand forecasting by better anticipating volatility. By leveraging improved methods with well-defined and systemically available incremental data inputs, it’s likely to demonstrate the value of these techniques to the organization, building broader visibility and engagement.
Additional objectives may include extending usage to a proven predictive use case for AI and ML, such as demand sensing or inventory optimization. Another option is to partner with data science experts within your organization or external, and draw on their knowledge and experience to identify the clearest opportunities to improve supply chain planning performance. Further, Intelligent Decision Platforms are continuing to gain market share in driving incremental decision augmentation leveraging AI and ML techniques.
How to understand the full data picture
AI and ML models require a substantial variety of pertinent data for training and testing, which can quickly become voluminous. You will need to build a broad dataset inclusive of internal data — from sales, marketing, customer service, quality and finance sources — to develop insight into cross-functional influences on demand.
This is a large undertaking and will require coordination with various departments. As a starting point, communicate upfront with key stakeholders about the objective, what is needed of them and how it will benefit their teams. Alignment across departments will aid in the process of gathering data.
These internal insights should be integrated with customers and supplier data where possible to gain greater visibility into demand patterns, inventories, sell-through and volumes. Additionally, leverage third-party data, either commercially or publicly available, including weather, social media, macroeconomic, financial disclosure or regulatory information.
By understanding the points in the process that are most reliant on human expertise, like where the expected marginal outcomes are not fully represented within historical data sets, you can target those opportunities to support better decisions, while retaining the capability for planners to override system automated decisions and suggestions.
Additionally, articulate the existing data latency, and determine what the cadence of the decisions that need to be made implies for the frequency of the data updates required to support optimal decision making. You can also assure data quality by focusing on data governance as an ongoing process supported by AI and ML techniques to ensure that the input data available to the planning tool is of good quality and ideally low latency — both initially and ongoing.
How to leverage AI as a tool for decision making
To make the most of AI and ML planning, organizations must apply the processes for decision-making. One of the key benefits of implementing AI and ML in planning applications is to move beyond real-time analytics and enable the move toward more autonomous planning functions.
Over half of planning organizations use data analytics to make the decisions and take action, according to Gartner’s Understanding Decision Making Models in Supply Chain Survey conducted from May through June 2022. The remaining organizations take a more limited approach, using data analytics to offer insight or generate recommendations. These teams need to progress from decision augmentation and support toward automation.
Organizations must define what automation means to their planning process, what decisions they are willing to automate, and how they will combine human domain knowledge and advanced analytics. These will automate simple and perhaps complicated decisions, and drive better augmentation and support for complex and chaotic decisions.
In order to gain maximum advantage from an AI and ML implementation, it is necessary to drive process transformation and organizational acceptance to leverage AI and ML tools to drive better decision augmentation and automation. This will require institutional knowledge, the right people, and culture to apply the technology within their particular supply chain organization and system architecture.
Begin by changing the ways people work in small, but meaningful ways such as implementing an autonomous process for simple decisions where cause and effect are well-established. Over time, you can work towards automating complicated decisions and adding an element of domain knowledge to improve prescriptive output and drive further decision augmentation.
Constant disruptions confront supply chains, signifying that the future will no longer reflect the past. Traditional planning methods that focus inward and look backward will no longer be sufficient. Supply chain planning leaders must recognize this and enhance their capabilities to include AI and ML to better manage disruptions through more agile, and data-driven, decision making.