Just as supply chain disruptions became the frequent subject of boardroom discussions in 2020, Generative AI quickly became the hot topic of 2023. After all, OpenAI’s ChatGPT reached 100 million users in the first two months, making it the fastest-growing consumer application adoption in history.
Supply chains are, to a certain extent, well suited for the applications of generative AI, given they function on and generate massive amounts of data. The variety and volume of data and the different types of data add additional complexity to an extremely complex real-world problem: how to optimize supply chain performance. And while use cases for generative AI in supply chains are expansive – including increased automation, demand forecasting, order processing and tracking, predictive maintenance of machinery, risk management, supplier management, and more – many also apply to predictive AI and have already been adopted and deployed at scale.
This piece outlines a few use cases that are especially well suited for generative AI in supply chains and offers some cautions that supply chain leaders should consider before making an investment.
Assisted Decision Making
The main purpose of AI and ML in supply chains is to ease the decision-making process, offering the promise of increased speed and quality. Predictive AI does this by providing predictions and forecasts that are more accurate, discovering new patterns not yet identified, and using very high volumes of relevant data. Generative AI can take this a step further by supporting various functional areas of supply chain management. For example, supply chain managers can use generative AI models to ask clarifying questions, request additional data, better understand influencing factors, and see the historical performance of decisions in similar scenarios. In short, generative AI makes the due diligence process that precedes decision-making significantly faster and easier for the user.
Moreover, based on underlying data and models, generative AI can analyze large amounts of structured and unstructured data, automatically generate various scenarios, and provide recommendations based on the presented options. This significantly reduces the non-value-added work that supply chain managers currently do and empowers them to spend more time making data-driven decisions and responding to market shifts faster.
A (Possible) Solution to the Supply Chain Management Talent Shortage
Over the past few years, enterprises have suffered from a shortage of supply chain talent because of planner burnout, attrition, and a steep learning curve for new hires due to the complex nature of the job function. Generative AI models can be tuned to enterprises’ standard operating procedures, business processes, workflows, and software documentation and then can respond to user queries with contextualized and relevant information. The conversational user interface commonly associated with generative AI makes it significantly easier to interact with a support system and affords the ability to refine the query, further accelerating the time it takes to find the right information.
Combining a generative AI-based learning and development system with generative AI-powered assisted decision-making can help accelerate the resolution of various change management issues. It can also accelerate ramp-up of new employees by reducing the training time and work experience requirements. More importantly, generative AI can empower people with disabilities by enhancing communication, improving cognition, reading and writing assistance, providing personal organization, and supporting ongoing learning and development.
While some fear that generative AI will lead to job losses over the coming years, others think it will level up work by removing repetitive tasks and making room for more strategic ones. In the meantime, it is predicted to solve today’s chronic supply chain and digital talent shortage. That’s why learning how to work with the technology is important.
Building the Digital Supply Chain Model
Supply chains need to be resilient and agile, which requires cross-enterprise visibility. The supply chain needs to “know” the entire network for visibility. However, building out the digital model of the entire n-tier supply chain network is often cost-prohibitive. Large enterprises have data spread across dozens or hundreds of systems, with most large enterprises managing more than 500 applications concurrently across ERPs, CRMs, PLMs, Procurement & Sourcing, Planning, WMS, TMS, and more. With all this complexity and fragmentation, it is extremely difficult to logically bring this disparate data together. This is compounded when organizations look beyond the first- or second-tier suppliers to where collecting data in a structured format is unlikely.
Generative AI models can process massive amounts of data, including structured (master data, transaction data, EDIs) and unstructured data (contracts, invoices, images scans), to identify patterns and context with limited pre-processing of data. Because generative AI models learn from patterns and use probability calculations (with some human intervention) to predict the next logical output, they can create a truer digital model of the n-tier supply network – faster and at scale – and optimize inter- and intra-company collaboration and visibility. This n-tier model can be further enriched to support ESG initiatives including but not limited to identifying conflict minerals, use of environmentally sensitive resources or areas, calculating carbon emissions of products and processes, and more.
Even though generative AI provides a significant opportunity for supply chain leaders to be innovative and create a strategic advantage, there are certain concerns and risks to consider.
Your Supply Chain is Unique
General uses of generative AI, like ChatGPT or Dall-E, are currently successful in addressing tasks that are broader in nature because the models are trained on massive amounts of publicly available data. To truly leverage the capabilities of generative AI for the enterprise supply chain, these models will need to be fine-tuned on the respective enterprise data and the context specific to your organization. In other words, you cannot use a generally trained model. The data management challenges like data quality, integration, and performance that hamper current transformation projects can also impact generative AI investments, leading to a time-intensive and costly exercise without the right data management solution already in place.
Generative AI is dependent on understanding patterns within the training data and if supply chain professionals have learned anything in the last three years it is that supply chains will continue to face new risks and unprecedented opportunities.
Security & Regulations
The basic requirement of generative AI models is access to vast amounts of training data to understand patterns and context. That said, the human-like interface of generative AI applications can lead to user impersonation, phishing, and other security concerns. While limited access to model training can lead to underperformance by the AI, granting unfettered access to supply chain data can lead to information security incidents where critical and sensitive information is made available to unauthorized users.
It is also unclear how various governments will choose to regulate generative AI in the future as adoption continues to grow and new applications of generative AI are discovered. Several AI experts have expressed concern about the risk posed by AI, asking governments to pause giant AI experiments until technology leaders and policymakers can establish rules and regulations to ensure safety.
Generative AI offers an abundance of improvement opportunities for those organizations that can tap into this technology and create a force multiplier for human ingenuity, creativity, and decision-making. That said, until there are models trained and explicitly designed for supply chain use cases, the best way to move forward is a balanced approach to generative AI investments.
Establishing proper guardrails will be prudent to ensure the AI serves up a set of optimized plans for each user to review and select from that are aligned with business processes and objectives. Businesses that combine “business playbooks” with generative AI will be best able to increase teams’ capacity to plan, decide, and execute while still optimizing desired business outcomes. Organizations should also consider a strong business case, security of data and users, and measurable business objectives before investing in new generative AI technology.