Generative AI is the topic of conversation these days and could be applied in many places and industries. While models of language are very interesting and useful in the office environment, it is the larger field of AI that is more relevant out of the office with respect to the physical processes of supply chain management. We will cover both angles in this article.
Large language models (LLMs) or Generative AI broadly offer increased productivity to desk workers by approximately 30-40%, according to a McKinsey study. These models will provide a simple language interface to other software and make workflow automation even more seamless. They will draft emails and reports. They summarize large documents, audio files, and videos. Most usefully, they represent an efficient way of searching through all the company’s records and not just getting a list of possible sources, but generating an actual answer to the posed question (this is called enterprise search).
As supply chain challenges are largely dominated by physical processes and imponderables, numerical accuracy and strict adherence to boundary conditions of schedule, weight, and price are required. LLMs are not designed to incorporate logical or causal reasoning or to do mathematical calculations. Such challenges require other AI methods, not LLMs, for their effective execution.
Enhanced Visibility and Transparency
Having blind spots in a supply chain is risky in modern competitive markets. Combining AI models with GPS, IoT sensors, and data from mapping services can provide granular, real-time visibility into the location, condition, and potential delays of shipments. Thus, AI combined with location intelligence ensures accurate routing and on-time deliveries, reducing inaccuracies and proactive mitigation of issues.
Accurately predicting demand prevents inventory shortages or overflows and takes into account historical data and market trends. AI enables optimized inventory levels across warehouse locations, which can be easily visualized to help with efficient resource allocation and cost savings.
Optimized Routes and Transportation
Google’s AI-powered solutions can create routing algorithms based on essential business parameters. The dynamic nature of mapping services analyzes real-time factors such as traffic congestion, weather conditions, and road closures to optimize delivery routes, slashing transportation time and fuel costs. This dynamic approach ensures the most efficient journey for each shipment, regardless of external circumstances. There is no limit to the mode of transportation as well. AI models can explore various options – road, rail, air – based on cost, speed, and environmental impact, enabling businesses to select the most efficient and sustainable combination for each shipment.
Combining many sources of information, AI can provide a systematic overview of all the business's carbon emissions of all scope levels. In addition to tracking this, AI can provide a digital twin that allows proactive decision-making toward a controlled reduction of emissions driving neutrality and other ESG goals.
Predictive Maintenance and Risk Management
AI acts as a fortune teller for your equipment. Analyzing sensor data from vehicles and warehouses predicts potential failures before they occur, allowing for preventative maintenance and minimizing costly downtime. Processing external data like weather forecasts, news feeds, and social media to anticipate potential disruptions – natural disasters, political unrest, etc. – allows businesses to proactively adapt their operations and minimize the impact of unforeseen events.
Having forecasted adverse events with numerical AI methods, generative AI can now help to organize the relief effort by analyzing manuals and spare part lists to decide who will do what; when and where to use which spare parts; and how to procure these in time. During the repair, assistance can be provided by AI using virtual reality, minimizing equipment downtime and maximizing chances of success.
Improved Customer Experience
No more wondering where your package is. Mapping systems provide customers with real-time order tracking, accurate ETAs, and proactive updates on any delays. This transparency builds trust, improves customer satisfaction, and fosters loyalty.
Navigating international services such as customs declarations can be a major obstacle for logistics companies in dealing with foreign languages, forms, and regulations. Generative AI can ingest large amounts of data - such as regulatory documents - in many languages and make sense of them. It is capable of translating text into other languages with uncommon accuracy and thus streamlining such bureaucratic processes.
Significant efforts are spent in resolving various problems arising in the supply chain, often requiring a customer hotline. More than half of hotline requests can be dealt with completely automatically by a well-integrated large language model. Not only does this obviate the typical frustrating menu options by allowing the customer to simply say what the problem is, but it can also answer or resolve these queries. It is a common observation that this automation shortens hotline calls by about 30% and improves net promoter scores by two points. Interestingly, it also improves job satisfaction for your employees staffing the hotline as they receive fewer annoyed customers venting their anger.
In conclusion, we see that there are many use cases that standard and generative AI can provide in the logistics and supply chain industry. The impact lies in the automation of repeatable tasks, thus saving either time, material, or cost. By saving time for the customer, you elevate the experience and thus raise revenue. Even though generative AI is mostly useful for knowledge workers, there are plenty of applications in and around the physical workflows of logistics supported by standard AI tools to provide forecasts and better data analysis.
Anthony Michael is Global Director, Location Intelligence Practice, Searce. Patrick Bangert is SVP of Data, Analytics and AI, Searce. Visit www.searce.com for more information.
This article originally appeared in the March/April, 2024 issue of PARCEL.
