Industries and media are abuzz about generative artificial intelligence (AI). These solutions are being described as the most significant tech launch since the iPhone. ChatGPT garnered 100 million users in just two months; shortly after, more generative AI solutions flooded the market, including Google’s Bard, Microsoft’s Bing AI, ChatSonic, YouChat and more. It’s no wonder this is one of ASCM’s Top 10 Supply Chain Trends for 2023.
Generative AI creates a novel output — such as text or an image — based on predefined rules and parameters. They also have the ability to interact with users in a conversation-like nature by answering follow-up questions, challenging incorrect premises, admitting mistakes and rejecting inappropriate requests. In essence, the platforms are competing with each other to offer the most human-like conversational experience.
AI has the potential to revolutionize business in all sorts of ways.
McKinsey says 2023 will be one of the most exciting years for AI yet. And the Gartner Hype Cycle predicts generative AI will be mainstream in non-supply-chain applications within the next two-to-five years. OpenAI, an AI research laboratory and the creator of ChatGPT, expects 80% of the U.S. workforce to have at least 10% of their work tasks affected, with higher-income jobs more likely to be influenced.
Although the technology certainly is intriguing, Gartner says it will take another decade before the supply chain field is significantly changed. Reece Hayden, senior analyst and AI lead at ABI Research, concurs: “ABI Research predicts three waves of generative AI use cases: process augmentation, product and service creation, and process automation. Most value-added supply chain use cases lie in the third wave of use cases, so the technology is simply not ready for widespread implementation.”
So why the delay in adopotion of AI in supply chain management?
The main thing slowing down generative AI in supply chain is the complexity of supply chain models. Furthermore, to be truly helpful, generative AI would have to be trained on company-specific models, rather than blanket best practices, which requires more technology maturity. It also demands quality end-to-end supply chain data, which can be difficult to collect and organize.
“ChatGPT can do what it does because it was trained with 570 [gigabytes of] data gathered from websites, books, Wikipedia, etc. across the internet — that’s more than 300 billion words,” writes Marko Pukkila, vice president, analyst and chief of research for Gartner Supply Chain. “How much good and clean data do you have that can be used for creating a complete digital image of your end-to-end supply network? Nobody really likes to take on master data initiatives, but it must be done.”
Early generative AI use cases in supply chain
Still, some low-risk use cases for generative AI are possible now, such as handling contracts and other documents. Flexport, a global logistics technology platform, adopted Scale Document AI to help extract data from a variety of unstructured logistics documents, including bills of lading, commercial invoices and arrival notices. By leveraging the tool, the company can extract data from these diverse types of documents at 95% accuracy in less than 60 seconds. This is a vast improvement from template-based, error-prone optical character recognition, which typically needs humans to correct its outputs. This accuracy and efficiency also help Flexport accurately report compliance data, minimize fines, reduce delays in delivering goods and reduce human workloads, it adds.
Yossi Sheffi, Elisha Gray II Professor of Engineering Systems at the Massachusetts Institute of Technology (MIT) and director of the MIT Center for Transportation and Logistics, adds that companies shouldn’t take too much stock in blanket predictions because organizations adopt technology at different rates. Those that are more advanced in their supply chain digitization efforts will have an easier time incorporating AI into their operations, he notes.
Other potential uses for generative AI in supply chain are mainly extensions of existing technology. Generative AI is good at analyzing historical data and using predictive analytics to forecast demand and supply. It also can help balance inventory to avoid stockouts and overstock, analyze risk, optimize delivery routes, and provide real-time updates to customers. Of course, as all of that is possible with existing technology, generative AI can certainly enhance these activities, but it probably won't revolutionize them.
Planning for generative AI in your supply chain
With these possibilities already on the table, companies should start evaluating the technology now to determine how it can make valuable contributions to their operations. “Supply chain leaders should encourage their people to experiment cautiously with ChatGPT,” Gartner’s Pukkila advises. He says “cautiously” because generative AI solutions offer limited data privacy. Therefore, only nonsensitive or fabricated information should be used for testing.
“This is an opportunity for us humans to learn a new way to interact with technology,” he continues. “It is not to be missed.”
Organizations that want to remain supply chain leaders should monitor this — and other disruptive technological innovations — and consider how it can improve their businesses and offer value. Here are six areas to explore:
1. Operations efficiency.
Generative AI can be used to draft an email, summarize a sales report or find the most important information in a spreadsheet. This can save supply chain professionals time in assessing various options for cost-effectiveness, analyzing supplier rates or evaluating supplier performance based on contract terms, for example.
The technology also has the capability to help predict demand, optimize inventory, analyze and identify risk, detect changes in production, identify bottlenecks, improve delivery routes in real time, and track shipments. Generative AI also can make some supply chain software more user-friendly.
Sanjeev Siotia, chief technology officer of Manhattan Associates, suggests that generative AI could allow a supply chain manager to query a warehouse management system, “Who are my best three pickers today?” or, “Who should I assign to the inbound dock?” to help manage operations. The conversational interface also could help a retail manager ask a warehouse management system what other distribution centers and stores have a particular item available for shipping to fulfill a customer request. The conversational exchange means that workers will need less technical training about the ins and outs of the software, and they can instead engage with the tool more naturally.
2. Customer service.
McKinsey points out that interaction labor, including customer service, has experienced less technology development than production and transaction labor, so the field is due for change. Chatbots can be an improvement over current chatbots and reduce the need for human customer service agents to step in and solve basic customer service issues. Generative AI’s advanced skills allow it to answer complex questions and understand unstructured information. When paired with an enterprise resource planning system, for example, a generative AI chatbot could search the system for information about specific orders to give customers detailed updates, among other services, without human input. Similarly, the availability of reliably helpful chatbots can allow companies to offer 24/7 customer service.
3. Customer relationships.
Supply chains’ global growth can be hampered by language barriers. Advanced chatbots can help supply chain professionals converse with their international partners by translating for them. In addition, this translation ability could be applied to customer service chatbots to offer customers assistance in more languages. “ChatGPT’s performance is also on par or better than most digital translation tools available today,” writes Emily Newton for Digital Commerce. “The fact that it can process natural language, including advanced technical terminology, gives it a major advantage over competitors. Its accessibility makes it ideal for supply chain applications, which rely on quick turnaround times and clear communication.”
4. Marketing and communications.
Generative AI certainly can be used to help with the more creative business operations, including creating marketing, social media and blog content (certainly not the ASCM Insights blog, though). Some of the earliest-reported uses were writing academic essays, much to the chagrin of educators, and creating text for marketing campaigns. By extension, the technology can be helpful for brainstorming for new product development.
5. Information technology.
Generative AI can offer customized fix recommendations as well as the code needed to address cybersecurity issues. One particular generative AI platform, DeepCode, is specifically trained to review software code for bugs. At the same time, hackers can use these tools to find and exploit vulnerabilities and develop malware. This means that IT professionals need to fight back quickly and efficiently, which this technology can also help with.
In addition, a survey by IDC found that a key challenge facing many organizations is effectively prioritizing and contextualizing the large amounts of data generated by cybersecurity alert systems and then identifying the key actions necessary to mitigate threats and vulnerabilities. Generative AI can analyze unstructured information and summarize it or organize it into a table for easier human understanding.
6. Sustainability.
Jacqueline Barbieri, founder and CEO of Whitespace, notes that one of generative AI’s most exciting and impactful capabilities would be assisting with sustainability. Generative AI could potentially be used to track supply chain greenhouse gas emissions and help with supply chain mapping to help companies see the impacts beyond tier-one suppliers. In the nearer future, generative AI can help with sustainability monitoring by analyzing text data from various sources, such as government reports or social media posts, to quickly extract relevant information about emissions. It also can learn from historical data to predict future emissions. This can help companies shrink their carbon footprints or help researchers and policymakers identify the countries or industries most likely to contribute emissions and work with them to reduce the environmental impacts.
Ironically, generative AI is very resource-intensive and currently not very eco-friendly itself, points out Thomas Kunnumpurath, vice president of systems engineering for Americas at Solace. “The extensive graphical processing units (GPUs) workload — now estimated to be upwards of 28,936 GPUs — required to train the ChatGPT model and process user queries incurs significant costs,” he explains. “The high-power consumption of GPUs contributes to energy waste, with reports from data scientists estimating ChatGPT’s daily carbon footprint to be 23.04 kilograms of carbon dioxide equivalent, which matches other large language models.”
What generative AI can’t do
Generative AI still has limitations. It holds strong potential to enhance human operations, but it definitely can't replace them. “Humans still outperform AI in a number of areas,” says Sheffi. “For example, humans are better at flexing with unexpected change, can draw on deep real-life experience, are able to make moral judgments when necessary and are more creative.”
Although the technology becomes smarter every day, its capabilities still are limited, particularly compared with human reasoning skills. For example, during the COVID-19 pandemic when a lack of historical precedents made it difficult for AI to offer accurate predictions, humans had to step in to forecast and act.
In addition, right now the technology is prone to hallucinations, or producing content with factual or reasoning errors, so an output needs human review before it can be used. Reece Hayden, senior analyst and AI lead at ABI Research, points out that using the incorrect information could have significant financial and operational implications, so companies should be careful using content from AI and limit its use to low-risk back-office activities at first.
However, because generative AI seems to be here to stay, supply chain professionals are at a point of adaptation, says Carm Taglienti, portfolio director and chief data officer of solutions integrator Insight Enterprises Inc. “My recommendation is to focus on learning what generative AI actually is, how it works and under what circumstances it can be used to make you more productive or efficient,” he says.
Alongside this knowledge, supply chain professionals should focus on analytics skills, Hayden advises. “Embedding generative AI into supply chain management will create additional access to high-quality data sources, so analytics will become an increasingly highly prized skill,” he explains.
Generative AI’s ability to make activities more efficient will ultimately require some employees to upskill to other activities, as was the case with the addition of robotics and other forms of automation. Still, AI is unlikely to replace supply chain jobs; hopefully, it will be a positive influence, making them more efficient and fulfilling.
