I was recently interviewed for a New York Times article about how Artificial Intelligence is being used to map supply chains for risk of forced labor. New rules in the United States require firms to prove their goods are made without forced labor, or they risk having them seized at the border. U.S. customs officials have already seized nearly a billion dollars' worth of shipments suspected of having ties to Xinjiang. Customers are demanding proof that products are genuine, ethically and environmentally sustainable, and conflict-free, forcing companies to explain where their products come from. Can AI really solve for costly interruptions to businesses operations and reputations? The answer is some and soon.
I've been chasing supply chain transparency for over a decade. I believe, to borrow a phrase, ‘the truth will set you free’ literally, figuratively, and economically. We simply cannot solve entrenched human rights issues without supply chain transparency. We cannot solve for climate change without tracing emissions in supply chains which account for on average 80% of a company's total emissions. Supply chain transparency is a quadruple win scenario: save time, money, people and planet. I founded FRDM with a mission to build a world where transparent supply chains give companies advantage in the marketplace. We have been building technology mapping risk in supply chains for the last five years and I don’t know of anyone more excited about the possibilities of AI to speed up this vision than us.
The reality is companies are struggling to trace their supply chains due to complexity and opacity. While artificial Intelligence is an exciting and promising advancement towards transparency, it is far from proven to deliver on the demand mapping companies suddenly need due to new trade enforcement laws like the Uyghur Forced Labor Prevention Act. I’ve spent the last ten years working with the people at the top of supply chains, and the people toiling at the bottom and both want the same thing, supply chain risk management. AI can’t solve all the problems it claims to, yet. AI currently can’t determine which fishing boat on Lake Volta has 8 year old boys who were purchased for $20 to work 17 hours a day, or the boys who are out with their father and uncle. AI currently cannot identify which mines supporting the production of cobalt in the Democratic Republic of Congo operate with children caked with toxic material on their faces, and which mines operate legally. AI is yet to surface purchase orders from a gravel company in Western China used to produce solar panels. AI can’t read the handwritten invoice written in malay located in the bottom drawer of your palm oil sub-supplier in Malaysia. AI doesn’t do these tasks, yet. And ‘yet’ is the operative word, because we are learning every hour what AI can deliver on seemingly intractable problems like these. While there is a lot AI cannot do yet, there are plenty of tasks for which AI is currently increasing the efficiency and effectiveness of supply chain risk mapping.
AI Is Powering Data Federation
As the business management axiom goes, what gets measured gets managed. Or, if you don’t measure your supply chain you won’t manage it. Supply chains have become increasingly complex since 2000, with many levels of suppliers involved in making products. Large companies like Procter & Gamble have nearly 50,000 direct suppliers, and orders of magnitude more sub-suppliers. Over 80% of supply chain data is unstructured, which is code for a hot mess. Less than 6% of companies have visibility beyond tier one suppliers. For most companies their supply chain data is a sock drawer where nothing matches. Covid exposed this (looking at you Wayfair dresser sitting on a container ship for 5 months). Disparate ERP (Enterprise Resource Planning) software, legacy IT systems with hand-cuffed data, multiple-language and multi-currency spreadsheets are just a few of the challenges procurement professions deal with on a daily basis. At FRDM we see some of the biggest companies in the world struggle with unstructured data issues, which is why we have begun to apply Large Language Modeling (LLM) on top of these enormously disparate data sets. LLM is saving months of work in data cleansing (matching), hydration (enhancing), and federation (connecting). Data federation is pulling 3rd party data sets against company spend data to generate insights into the proximity and salience of risk. This is a process we’ve always handled for our customers, but LLM is speeding up the process up to 400%.
AI Detecting Risk Signals
Where there is smoke, there is fire. While not all fires are directly connected to your business, they can still burn. Let’s say one of your suppliers is doing business with a sub-supplier who is doing business with another supplier that is sanctioned due to forced labor. If this connection is detected by Customs and Border Protection your next shipment may be detained at port, and you will have 30 days to prove no forced labor was involved. Applying learning models to trade data and sub-tier supplier relationships is allowing companies to detect potentially harmful patterns early in their supply chain. Just like AI is being deployed to predict logistical risks caused by weather patterns, it can also be used to predict human risks like cross referencing climate data projections with human migration predictions. Where there is migration there is exploitation. The data shows this out.
AI Predicting Risk
Elon Musk talks about how designing a car or rocket is the easy part, but designing the supply chain to support production is what’s hard. Once a supply chain is designed, you want to protect it at all costs by monitoring potential disruptions. AI is beginning to be used to predict human rights risks in supply chains. The exploitation of workers and other human rights abuses can occur throughout complex global supply chains, creating reputational, legal, and ethical risks for companies. Exploitation is a behavior, and therefore hard to track scientifically. You have to track it referentially. That’s why AI is so critical to analyze vast amounts of data from various sources, including supplier records, social media, news articles, and other public data, to identify patterns and indicators of potential human rights violations. AI can also learn from past incidents and predict the likelihood of future risks, allowing companies to take preventive action. AI can also help identify and monitor high-risk suppliers and locations, enabling companies to focus their efforts on mitigating human rights risks where they are most significant.
AI Supporting Risk Mitigation
According to a study by Deloitte, AI-powered supplier risk management can help organizations reduce the time it takes to evaluate supplier risk by up to 63%, enabling them to make more informed decisions and improve their supply chain resilience. At FRDM we are seeing even faster evaluation periods. AI can also be used to prioritize mitigation workflows used by trade compliance and procurement teams. There is no such thing as a perfectly ethical supply chain. AI can prioritize tasks based on their potential impact on business operations, financial costs, and reputational risks. This allows companies to focus their resources on the most critical risks and take proactive measures to mitigate them, updating risk assessments in real-time, providing companies with continuous monitoring and feedback to ensure they stay ahead of potential threats.
The reality is that AI is solving a lot of problems very quickly, and it’s only going to get better. We created FRDM to build a supply chain transparency movement allowing businesses and consumers to buy with their values. Artificial intelligence is speeding this movement up by decades. To learn more about this movement and how your company can join visit www.frdm.co.