Shahzaib Khan discusses Resilinc’s journey toward AI-driven supply chain excellence
From its early inception in the 60s’ to its increasingly available public access, artificial intelligence (AI) has come a long way. As with any technology that innovates throughout the decades, AI has many milestones and historical roots—from its use in some of the first early natural language processing (NLP) programs (circa Daniel Bobrow’s STUDENT NLP) to IBM Watson beating Ken Jennings on Jeopardy in 2011, to today’s use of Alexa, Siri, ChatGPT, Google Bard, Microsoft’s CoPilot and more. In some cases, it feels like AI is just starting to take off.
In reality, supply chain AI has been helping to address the complexities of global supply chains for over the last decade. (Understand the nuances of multi-tier mapping vs AI mapping here). Yet, numerous global supply chains today still struggle with mapping their global suppliers. For example, in the life sciences industry, biopharma companies increasingly rely on contract manufacturing and development organizations (CDMOs) to address biologics manufacturing production and capacity demand. While the benefits of using CDMOs are multifold, the scenarios also create deeper, more complex supplier networks with worse supplier visibility and greater difficulty in predicting and preventing supply chain disruptions. Enter the future of AI in supply chain.
In this new multi-part blog series, we address the possibilities of supply chain AI and how new capabilities and machine learning algorithms are revolutionizing supply chain risk management (SCRM). We will define: What are the nuances between predictive vs generative AI for supply chain? Are there different models of AI for different supply chain use cases? How is AI impacting different industries challenged by labor issues, recent geopolitical tensions, commodity restrictions, import/export sanctions and more.
In the first part of our AI blog series, we sat down with Resilinc’s Associate Vice President of Product Management, Shahzaib Khan to understand how far we’ve come with supply chain AI and address how global leaders and their suppliers can harness the power of AI for greater supply chain resiliency.
We’ve redefined the use of predictive AI in supply chain for the past decade to help global leaders address gaps in their SCRM journey—how far have we come?
Shahzaib: We were the first company in 2011-2012 to create a supply chain disruption alert system (EventWatchAI) that could monitor time-sensitive global events for faster response and mitigation of supply chain disruptions, 24/7, from over 104 million sources in 108 languages. This generates massive amounts of unstructured data which must be classified and sent to customers promptly. Over time, as we added more customers, industries, and suppliers, we quickly realized that this system could only be scaled with robust technology. That’s when we started looking into transformers.
We were among the early adopters of the BERT model (based on transformers) for natural language processing (NLP)—which 5 years ago was at the forefront of AI research. BERT revolutionized understanding and processing textual data. EventWatchAI employs classification models that fall under the category of NLP. Our initial use of BERT allowed us to improve our text-understanding capabilities significantly. We used it to enhance our search and recommendation systems, providing our users with more accurate results and recommendations than ever before.
Today, we monitor and alert customers of 40 different event disruption types—and receive about 8 million rows of data daily. Without a proper robust AI-based system, dealing with this vast amount of data would be impossible.
Today’s global leaders and their suppliers face even more supply chain disruptions than ever before—how are we solving for their challenges, present and future?
Shahzaib: Following our launch of EventWatchAI, our customers brought forward more use cases. As the AI landscape evolved, we recognized the need to stay ahead of the curve. In 2021, we identified and swiftly implemented RoBERTa into our systems, enabling us to enhance our event classification capabilities further. In 2023, we further recognized the potential of large language models (LLMs) for event classification and sentiment analysis. We saw that our clients, particularly those in the supply chain industry, could greatly benefit from our ability to scan various textual sources to predict and identify relevant events and supply chain disruptions that could negatively impact their business. We innovated again to include LLM for end-to-end execution on event identification to event publication.
Now, we leverage AI in tracking purchase orders (POs) and commodities, autonomous mapping, risk-scoring models, and simulators. Across these different solutions, we applied various AI models, including predictive models, classification models, clustering models, and recommendation models. These different models were trained on several data sets: historical data provided by customers and third-party data, as well as Resilinc’s 12+ year wealth of supply chain data.
How has autonomous AI mapping in supply chain changed over the years?
Shahzaib: With the help of AI in autonomous mapping, we process unstructured data to identify relationships between companies to create a map of every company’s supply chain several tiers deep. For example, imagine we recognize that suppliers A and B have a trade relationship through autonomous mapping. Then, we look at where each company is manufacturing, housing, assembling, and testing.
We try to verify and extract those attributes from 104 million sources of publicly available data that we monitor for supply chain disruptions to increase visibility and mitigate risk from changing scenarios, such as new industry regulations. In fact, autonomous AI mapping is more important now for compliance. Companies are getting sanctioned as a result of greater supply chain regulatory legislation like UFLPA, geopolitical conflicts, and increased sustainability/ESG initiatives. Because of this, our customers need quick visibility into where sanctioned companies sit within their supply chain.
To truly embrace AI for SCRM, what do users need to first address?
Shahzaib: Many global supply chain leaders still have a big black hole beyond Tier-1 suppliers. They have no idea who their Tier-2 and Tier-3 suppliers are or where they are located. While autonomous mapping can help organizations gain visibility, a hybrid approach of multi-tier mapping and autonomous AI mapping can provide not only the right amount but quality of data necessary (learn more here).
Why specifically AI for POs? How does it work?
Shahzaib: We developed AI for POs because our customers had no visibility into POs. Disruptions delaying their POs were always a surprise, so they were always in a reactive state instead of a proactive state. To remedy this, we created AI to monitor disruptions several tiers deep. This AI monitors disruptions happening at logistics waypoints and connects the dots back to historic disruptions and delays. Then, it provides a future delay prediction on active POs.
To achieve this, we utilized historical purchase order data from customers and Resilinc’s 12+ years of supply chain data. Trained on this data, our machine learning models identified why historical purchase orders were delayed and for how long. The AI that goes into our specific PO delay prediction offering is a type of predictive model called a regression model, which is utilized to predict continuous numerical values based on input features (stay tuned for AI blog 2: Predictive vs Generative AI).
How will AI play a role in SCRM in the future?
Shahzaib: Utilizing AI can transform supply chain risk management in several ways, making it more efficient, proactive, and responsive to potential disruptions. Yes, remaining proactive in one’s SCRM journey is critical today, but competitive differentiation also matters. Today’s organizations with global supply chains and multi-tier suppliers want to take advantage of offering their suppliers with capabilities around global events or situations that their competitors may not yet be aware of. At Resilinc, we are innovating for the future via research and discovery of generative AI and reinforcement learning and their impacts across our operations, new risk model developments, product marketing, code development, internal data efficiency operations, and more.
Having a robust SCRM program can also support other cross-organization initiatives and departments as well. For example, compliance, procurement, IT departments, and logistics—there are a lot of areas of a business that can equally benefit from a holistic SCRM program backed by AI and machine learning as well. By leveraging AI, businesses can build more resilient and agile supply chains, reducing the impact of risks and improving overall efficiency.
Stay tuned for part 2 of our AI blog series. Looking to continue your journey toward AI in supply chain now? Learn more about Resilinc’s Autonomous AI Mapping here.