How is AI Transforming Supply Chain Operations?

Even in 2025, it’s not unusual for warehouse workers to pull information from multiple systems using spreadsheets, whiteboards, emails, and other largely manual tools to gather the data, says Keith Moore, CEO of AutoScheduler, which offers a warehouse orchestration platform. All the work required to manage the data leaves less time to draw insight from it.

If businesses instead pull the data into a single repository and leverage artificial intelligence (AI) solutions, they can spend less time keeping their warehouses going day-to-day, and gain greater intelligence to drive decisions, Moore says.

While the promise of AI has been touted for years, companies that have implemented AI solutions are now starting to see returns on their investments through operational and labor efficiencies, better tracking of capital assets, and more accurate forecasting and insightful decisions, among other benefits.

“AI is not a science project anymore,” Moore says. And while AI solutions haven’t yet hit the majority of companies, he predicts that they’ll be mainstream in about five years.

Companies that move first and fast to leverage the power of AI can gain an edge. Here are a few companies doing just that.

Pepsico’s Warehouse Efficiency Bubbles Up

PepsiCo leverages AI to enhance warehouse labor management.

As operations at PepsiCo’s warehouses grew more complex and leveraged different technologies, the beverage company increasingly relied on its experienced employees to determine how best to move products from various areas within the warehouses to their final destinations, said Peter Hall warehouse orchestration senior manager, in a recent podcast with AutoScheduler.

These employees had to research each load and then determine if they had the right products and people in place to complete them.

When the pandemic hit, the average tenure and experience of warehouse employees dropped. A warehouse orchestration system from AutoScheduler now helps PepsiCo determine, among other information, the number of employees and the amount of time and space needed to activate a planned number of loads. Sites that have implemented the software are averaging about a 12% increase in moves per hour, Hall said.

SGWS drinks in Forecast Accuracy

Southern Glazer’s taps into Amazon’s SageMaker AI solution to generate demand forecasts.

Southern Glazer’s Wine & Spirits (SGWS) distributes beverage alcohol, with operations in 47 U.S. markets, as well as Canada and the Caribbean. Each year, it makes more than six million deliveries, distributing more than 11,000 supplier brands to both national retailers and local restaurants.

Southern Glazer’s has done a great deal of work leveraging AI to refine its sales forecasts, says Diego Fonseca, vice president, supply chain and logistics for SGWS. Not only does the company have vast quantities of sales data—some extends back one century—but planners incorporate into their forecasts the intelligence they receive from the sales, trade development, and commercial operations areas. Manually cleaning and working with this volume of data is tremendously time consuming.

Today, planners can feed this information into Amazon’s SageMaker AI solution, which can generate a demand forecast that considers historical data, current promotional events, and seasonality, among other factors. “Artificial intelligence does well with long histories and databases,” Fonseca says.

Similarly, to determine when to place orders, SGWS’s replenishment team continually looks at forecasted demand, current inventory levels, and supplier lead times, while factoring in potential disruptions such as port strikes, as well as suggested order quantities. They also work with suppliers to account for product availability.

The AI solution can quickly identify different patterns and use machine learning to generate replenishment models. Planners can spend less time fine-tuning the more predictable business lines and instead can focus on improving the accuracy of more complex forecasts, such as those for spirits.

SGWS launched its AI program in spring 2024. “Implementing it is a huge lift from the IT side,” Fonseca says. Establishing the data infrastructure requires defining data quality and sourcing, and deciding how to create the data models, among other tasks.

A partner with experience in AI initiatives can provide valuable insight, Fonseca says. At the same time, it’s a mistake to simply hand the effort to an outsource partner. The company needs to confirm that the data, infrastructure, and algorithms are solid and work as intended. Fonseca says his team worked closely with data scientists to continually evaluate the results. SGWS has continued using Blue Yonder for its ERP system, creating an integration between it and SageMaker.

Initially, about one-quarter of SGWS planners worked with the AI solution. That’s since grown to about 55% and employees see how it helps boost productivity,

With the assistance of AI, SGWS’s 2024 forecasts were consistently about six points better than they had been, Fonseca says. The AI forecasts gain accuracy as the forecasted dates move closer to actual dates. This helps with purchases of domestic products, where lead times tend to be shorter than they are for cross-border purchases.

SGWS has also begun using AI to convert thousands of carrier invoices into usable data. Given the volume of invoices from different carriers, the company had no practical, manual way to track this expense and identify ways to reduce it. With AI, Fonseca and his colleagues can quickly separate spending by attributes such as fuel, provider, or region and then zero in on savings opportunities.

These initiatives are part of Lumina, a 10-year strategy and vision for unlocking the full potential of SGWS’s supply chain and operations engine to deliver long-term value. It includes substantial investments in artificial intelligence.

Werner Streamlines Trailer Recovery

Werner Enterprises integrates AI across various facets of its operations to enhance efficiency and safety.

Prompted by the increasing challenge of unauthorized carrier use and missing equipment, Werner Enterprises implemented GenLogs, an AI-powered solution to streamline the recovery of missing trailers, in mid-2024. “These incidents created inefficiencies, increased costs and reduced visibility into equipment utilization,” says Daragh Mahon, executive vice president and CIO with the 3PL. Werner needed a solution that would provide real-time tracking and recovery capabilities.

GenLogs monitors equipment through camera systems deployed over the road, identifying trailers that have been flagged as missing, or those with malfunctioning geolocator devices. “We eliminated the guesswork previously involved in tracking down missing equipment,” Mahon says.

GenLogs reduces the time required to locate missing trailers from days or even weeks to mere hours, Mahon says. It also enables Werner to identify instances of unauthorized trailer use, so the company can take corrective action swiftly. With concrete proof of unauthorized use, Werner has been able to settle numerous claims, which translates to better resource allocation, reduced downtime and enhanced operational control, ultimately driving a more efficient supply chain.

Werner is also leveraging AI insights to analyze lane traffic patterns. This insight may further optimize logistics operations.

Standard Logistics Optimizes Loads

Standard Logistics uses Optimal Dynamics’ Decision Automation Platform to optimize loads.

Standard Logistics began using an AI-based optimization engine from Optimal Dynamics in 2021 to determine which of the loads offered to its fleet it should accept, and which it should pass to its brokerage solution.

“We had too many loads to choose from and didn’t have a proper optimization mechanism to ensure we picked the right loads for our fleet,” says Volker Bargenda, president of the 3PL.

Humans can’t process the thousands of variables needed to make optimized freight planning decisions, let alone account for the inherent uncertainty that exists throughout a carrier’s network, says Erica Frank, vice president of marketing with Optimal Dynamics. AI technology can determine the optimal loads based on multiple factors including service commitments, revenue and backhaul potential, among others.

Once a load is accepted, the solution also helps Standard Logistics identify which driver to assign to which load, using information from the transportation management system to make this decision. This includes the drivers’ available hours of service, customer commitments that may impact prioritization, and load profitability. The solution balances revenue potential with operational efficiency.

Standard Logistics sees a strong correlation between increases in revenue per driver as the company more consistently uses Optimal Dynamics’ AI solution to accept and dispatch loads, Bargenda says.

CJ Logistics Enhances Employee Safety

OneTrack’s AI solution helps CJ Logistics manage employee safety.

Employee safety is key at CJ Logistics America. The goal “is to make sure that all employees that come to work leave the same way,” says Laura Adams, senior vice president, continuous improvement, technology, engineering, solutions and business process integration.

That’s not always easy for warehouse workers. “Warehouses are massive buildings, with lots of square footage, lots of people, and lots of machinery crossing paths,” says Evan Stinson, director of marketing with OneTrack, which offers a warehouse operating system.

Because of their size and complexity, it’s difficult to continuously monitor every operation within a warehouse. Even security cameras can miss dead zones, or they might produce images that are too grainy to be useful.

In 2018, CJ Logistics partnered with OneTrack to better understand the actions happening in its warehouses and work to prevent accidents and injuries. Cameras mounted on equipment within the warehouse capture employees’ actions.

OneTrack’s AI solution can identify anomalous events, such as a task that takes longer than typical or a forklift that bumps into a rack of shelving. Obtaining this detail without AI would require people watching cameras all day long, Stinson says.

Supervisors can receive shift summaries highlighting employees who could benefit from additional training, along with video context to help provide instruction.

With this information, supervisors can coach their employees in a focused way, reducing potential and actual safety issues, Adams says.

The solution may also identify safety issues that regularly occur on a certain shift. This could mean the supervisor would benefit from additional coaching.

Implementing a solution like this can prompt privacy concerns. A company implementing it needs to let employees know it’s a tool to make it easier for them to remain on task, Adams says.

Moreover, the OneTrack solution looks at behaviors and doesn’t capture biometrics or employ facial recognition, Stinson says. The company works with each state to ensure it complies with applicable privacy regulations.

At CJ Logistics, the OneTrack solution has cut potential safety events by nearly three-quarters, while boosting units per hour by an average of 11%. It also cut product damage by 60%.

CJ Logistics is currently working on a project to upload the CAD warehouse layout, which would tie in with the OneTrack camera data. This could aid in several ways, such as optimizing travel paths, reducing congestion areas, and improving zoning and slotting.

DISA Documents results

DISA Global Solutions offers employee screening services, such as background screening, drug testing, and fleet and driver compliance. The solutions are “highly document-intensive,” and DISA processes millions of forms and records annually, says Steven Spencer, senior vice president of strategy.

DISA implemented an AI solution in 2024 to manage a growing volume of customer orders, while ensuring accuracy and efficiency. Artificial intelligence enables the company to scale faster without sacrificing quality.

While paper-pushing has been the butt of many jokes, documents are critical for a range of experiences important to both individuals and businesses, such as homebuyers securing mortgages or importers waiting for inspection reports. Hypercell, the AI-native platform from HyperScience, reads, understands, and processes documents with 99.5% accuracy.

It continually learns and adapts to an organization’s documents, enabling the automation of various processes, such as contract review and claims decisioning.

DISA is leveraging artificial intelligence primarily to automate document sorting and processing, such as extracting information from driver qualification files.

“Our AI-driven automation has resulted in a 99.5% accuracy rate in document classification and the vast majority of classification is now automated,” Spencer says. “This has saved time and money and enhanced our ability to scale.”

7 Steps to Successful AI Implementation

These guidelines can help organizations lead AI implementations that achieve the benefits they’re looking for.

1. Identify the right problems. A good AI project is one that a person could solve if they had all the data and all the time in the world. “If a person could do it, an AI can probably figure it out,” says Keith Moore, CEO of AutoScheduler. Conversely, if the necessary data isn’t available, AI might not be the best solution.

2. Prioritize change management. Technology alone isn’t enough. Employees may resist AI if they fear job cuts. Successful AI implementations show how the solution presents an opportunity to redefine roles and elevate employees into more strategic positions.

3. Insist on transparency from the AI provider. The provider should go beyond buzzwords and dive into the core technology and show how it directly addresses the company’s challenge.

4. Push your vendors. If existing solutions don’t meet your needs, challenge vendors to adapt or enhance their offerings to align with your requirements.

5. Implement quickly. The sooner a solution is operational, the faster you can realize the benefits and refine the approach.

6. Keep an open mind. Artificial intelligence can drive value in areas you may not have initially anticipated, Mahon says.

7. Remain flexible. Because the AI landscape is evolving, it’s crucial to be willing to adapt. “Flexibility has allowed us to refine processes and adjust to new developments in AI technology,” says Spencer of DISA, adding that AI is not a one-time fix but a continuous improvement journey.

How Has AI Helped in Your Supply Chain Operations?

AI Superpowers Warehouse Robotics

Agility Robotics

Agility Robotics uses NVIDIA’s AI acceleration platform for real-time perception and reinforcement-learned controllers onboard its humanoid robot Digit. The platform allows humanoid robots to host robust models that can process sensor data and make decisions in real time to improve interactions with humans in dynamic environments.

Agility Robotics is expanding adoption of NVIDIA Isaac Sim and NVIDIA Isaac Lab robot simulation and learning frameworks to train and test behaviors on Digit. Using these frameworks, which have AI models trained across billions of instances, Digit can improve via reinforcement learning in areas such as stability and step recovery.

Amazon Robotics

Amazon’s robotic system, Sequoia, uses AI, robotics, and computer vision systems to consolidate inventory and free up storage at its fulfillment center in Shreveport, Louisiana. Speeding up order transactions, it works by relying on mobile robots to transport inventory directly to a containerized storage system or to an employee picking out items for a customer order.

Ambi Robotics

Ambi’s AI-powered AmbiStack automates pallet packing by analyzing, tracking, picking, and placing packages using over 200,000 hours of warehouse data. The company’s new foundation model, PRIME-1, was pre-trained on 20 million images from 150,000 hours of robotic sorting operations. PRIME-1 enhances AI-powered robotic performance by enabling advanced 3D reasoning for tasks like depth estimation and picking. This technology improves efficiency, reliability, and adaptability in warehouse automation.

Boston Dynamics

Boston Dynamics is advancing AI capabilities for humanoid robots through its collaboration with NVIDIA. Its Atlas robot leverages the NVIDIA Jetson Thor platform to run complex AI models for whole-body control and manipulation. Using Isaac Lab, Boston Dynamics is developing advanced AI policies for dexterity and locomotion in virtual environments. Additionally, the company is integrating new AI capabilities into Spot, its quadruped robot, and Orbit, its fleet management software.

Dematic

Dematic uses AI to power the software that coordinates warehouse operations—adapting workflows, predicting demand, and guiding autonomous robotics. This intelligence enables faster, more responsive decision-making to help customers meet shifting supply chain demands.

OSARO

OSARO, a provider of machine-learning-enabled robotics for high-volume fulfillment centers, launched OSARO AutoModel, which enables robots to automatically learn and adapt to new items, processes, and workflows in real time, minimizing downtime. It accelerates the introduction and onboarding of new SKUs and increases robot productivity in kitting, piece-picking, and autobagging.

Lab0

Already deployed at Dollar Tree, Lab0’s AI-powered robotic system uses two autonomous arms in tandem with sliders to pick and unload packages. Designed for high-volume distribution centers, the automation system uses NVIDIA Isaac Sim, a robotics simulation platform built on NVIDIA Omniverse, and NVIDIA Isaac Lab, a framework for simulation and reinforcement learning. These technologies enhance real-time perception, decision-making, and robotic precision.

ZenaTech

ZenaTech’s subsidiary, ZenaDrone, is developing an AI-powered drone swarm – a coordinated group of autonomous drones that communicate and work together using AI and real-time data sharing to perform tasks collaboratively without direct human control – application for inventory management and security. This system uses autonomous IQ Nano drones that communicate and collaborate in real time without direct human control. Equipped with integrated sensors, high-quality cameras, and AI-driven data analysis, the IQ Nano enables efficient automation. With at least 20 minutes of flight time and automatic battery recharging, it ensures stability, safety, and obstacle avoidance.