By Are Traasdahl and Dag Liodden
With the rapid adoption of generative AI technology like ChatGPT, there’s no question that we’re in a breakthrough moment in history. In the months and years ahead, thousands of companies will build on top of large language models and explore their many applications. One promising application is using GPT agents to analyze data, which has the potential to revolutionize the global retail supply chain.
Why? In short, data is everything in retail. Information about what consumers want, how much product is available, and what’s getting shipped where is critical to helping retailers and suppliers meet consumer demand and drive profits. But data about the supply chain has sometimes been challenging to get, let alone drive meaningful insights from.
Thankfully, that’s starting to change. And with the right data inputs, GPT-powered AI has the potential to transform the retail industry, leading to stocked shelves, happier consumers, less waste, and higher profits.
Enabling data across the supply chain
There are endless opportunities to apply large language models like GPT-4 to retail, but three main categories come to mind:
Democratization of data: Data is a powerful tool in CPG, but it’s often siloed in organizations. GPT models can change that by helping nontechnical users analyze data to answer key business questions, relying on natural language instead of SQL or data scientists. For instance, "A conversation with a GPT analytics bot might start with a question like "What was the average sales velocity of my peanut butter product in Chicago last year?". The bot quickly replies and notes that the velocity is slightly down year over year. The real value of AI-supported analytics arrives when the user asks, "Why is it down?" The bot does a root cause analysis on the fly.
Proactive monitoring: Data may be the answer to important business questions, but what if you don’t know what question to ask? AI can proactively monitor sales and inventory data across the supply chain and alert retailers and suppliers to important events and anomalies if sales drop more than 10% at a certain retailer. While machine learning can do this today, it’s labor-intensive to set up and requires ML engineers to train the model on what you’re looking for. With large language models, AI can learn on its own what is normal and what is an anomaly, including patterns we haven’t yet thought about. For example, during the pandemic’s early stages, AI could have caught changing consumer behaviors and recommended increasing production of toilet paper or pantry staples before we faced empty shelves.
Autonomous action: The next step AI can take is not only detecting a pattern and making a recommendation, but automatically taking action based on its findings. Countless events are happening every day across thousands of retail stores, warehouses, or manufacturing facilities, and humans in retail today must stick to the priorities they can realistically implement on their own. But with AI, they can monitor and take action on any of these events, adding up to real wins. If AI detected rising temperatures during a heat wave, for instance, it could automatically allocate additional ice cream inventory toward the hottest stores.
Real-world applications already underway
Brands and retailers are already employing ChatGPT to serve shoppers better and streamline operations – using AI Chatbots in customer service, for example. Further advancements in AI could make these interactions more personalized and valuable. Retailers and brands alike strive to build relationships with consumers through personalized recommendations, from product suggestions to recipe ideas. With the vast amounts of customer data retailers collect through their loyalty programs, the ability to process and learn from that data will be a valuable tool for marketers.
On the operations side, Walmart is already using AI to negotiate with suppliers on the price of wholesale goods based on supply chain data. Another opportunity we are already leveraging at Crisp is to optimize supply chain management by detecting retail voids: situations where a product should be selling at a given store but isn’t due to an out-of-stock, misplaced or damaged merchandise, or other error. AI models can learn normal sales patterns at a store, identify an anomaly, detect a void, and alert the supplier or retailer – solving one of their biggest pain points in keeping shelves stocked and products selling.
AI can also be used to learn from ingredient lists or consumer preferences and suggest new products. Tech-savvy brand NotCo has an AI-powered product development engine, Giuseppe, that can rapidly develop plant-based analog equivalents to traditional dairy and meat products. Giuseppe collects data on plant molecules that are found in dairy – including cabbage and pineapple – then puts them together to deliver a product that is remarkably similar in taste and texture to cow’s milk.
Concerns and caveats
With all the possibilities around AI, serious considerations must be addressed before realizing its full potential in the supply chain, including:
- Start with good data: Retail data needs to be clean, structured, and accurate for AI to provide valuable analysis. Retail data platforms and ETL solutions can help by feeding clean data from retailers, distributors, e-commerce sites into AI-enabled tools.
- Context is key: Current GPT models can only take in a limited amount of context, and brands with dozens of products across thousands of retail stores can easily exceed that limit today. Retail will require GPT models that can process a higher volume of input.
- Consider the ethics: Ethical concerns abound with AI, and the supply chain is no different – especially if decisions are being made about where food is allocated. For example, in the case of a pandemic-related food shortage, we could not trust AI today to decide how and where to allocate inventory without considering social good and public health.
While there is plenty more work to do, we’re already seeing AI’s potential to do what we alone could not: create a supply chain that is more responsive, agile, profitable, and sustainable in meeting the needs of consumers.
About the authors:
Are Traasdahl is Co-Founder and CEO at Crisp, an open data platform that harmonizes and normalizes retail data to increase profitability and reduce waste.
Dag Liodden is Co-Founder and Chief Product Officer at Crisp, an open data platform that harmonizes and normalizes retail data to increase profitability and reduce waste.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.