Hype vs. Reality: The Promise of AI in Supply Chains

Some people say artificial intelligence (AI) will produce an era of unprecedented prosperity. Some fear it will wipe human life off the earth. Between those two extremes lie myriad claims about the benefits AI can bring.

While some of those claims are hype, this still-evolving technology offers a great deal of real value in the world today, including in the supply chain.

Supply chain professionals are approaching AI with caution. Just 25% of organizations are using new applications or insights from AI, finds a survey of 145 shippers conducted by Odyssey Logistics in 2024 (see chart at left).

When marketing firm RRD surveyed more than 300 supply chain decision-makers, it found that while AI is gaining interest, it’s attracting less investment than several other technologies, such as real-time visibility and the Internet of Things.

Companies that have embraced AI, though, apply it to some of their most central supply chain functions, RRD says: 59% use it to support supply forecasting, 56% for visibility and tracking, and 56% to help optimize their operations.

From Suggestion to Transaction

Among the supply chain solutions built on AI, many are designed to quickly extract insights from large, ever-changing databases.

For instance, AutoScheduler.AI, a technology provider in Austin, Texas, offers an AI platform that synthesizes data from a warehouse management system, enterprise resource planning system, transportation management system, and/or other solutions to orchestrate warehouse operations.

“We are continuously planning, dynamically prioritizing what needs to happen, finding where the bottlenecks will be, and what is the best way to run each facility to maximize throughput and minimize cost,” says Keith Moore, the company’s CEO.

A skilled warehouse manager can achieve something similar, but experienced talent isn’t always available. Also, humans can’t respond to changing conditions as fast as a well-trained AI solution.

“In some cases, every five to 10 minutes the AI says, ‘The world has changed. Here is what to do now,’” Moore says. As AI attends to the details, it frees humans to manage staff and deal with exceptions.

It also automates communications among various supply chain functions. “We can tell the transportation team what to communicate to their carriers around load readiness for every individual trip,” Moore says. “We can tell the manufacturing team how the warehouse will support them with raw materials, as well as the outflow of finished goods.”

UP.Labs in Santa Monica, California, has worked with Porsche to launch a new company, called Sensigo, that applies AI to improve auto diagnostics and repairs. “We are able to bring together dealer and OEM [original equipment manufacturer] data to bring the most likely fix to the service technician,” says Katelyn Foley, president of UP.Labs. The AI also detects patterns that point to problems with certain auto components, which Porsche can use to provide feedback to vendors.

As a “venture lab,” UP.Labs works with transportation companies to create startup firms to address those companies’ biggest challenges. These partnerships focus on technologies driven by AI and machine learning.

Computer maker Dell uses an AI decision intelligence platform from Aera Technology to automate decision-making across its supply chain.

In 2023, Dell started using Aera’s solution to help manage finished goods in the United States, said Sasha Pailet Koff, Dell’s senior vice president, digital supply chain, speaking at the AeraHUB 2024 conference.

These days, Dell often lets the technology not just make suggestions to humans, but also execute decisions automatically. And Dell has applied Aera’s technology to more processes.

“We’ve added a demand-sensing skill, and one that’s looking at aging inventory,” Koff said.

Also, Dell uses Aera’s AI to help its supply chain organization and its commercial organization collaborate more closely.

Nearly 90% of recommendations from AI are accepted, with 77% automatically written back to source systems. “That represents roughly 30,000 of what we call truly significant decisions that theoretically would have been in the hands of human capital,” Koff said.

Risk Management

Supply chain platform Inspectorio applies AI to procurement and production. The company’s offerings include solutions such as Quality Risk Management, Responsible Sourcing and Compliance, and Traceability and Transparency.

Like Aera Technology, Inspectorio can analyze large volumes of operational data to help human operators make decisions, or it can execute on those decisions automatically.

It can also head off problems before they occur. “One of our capabilities is predicting what factory, or what factory’s product, will cause a failure or a defect,” says Chirag Patel, Inspectorio’s CEO.

“We can make recommendations on either not using that factory, or inspecting the factory, or asking them to fix certain procedures and processes so we can avoid the defect,” he adds.

Inspectorio trains the AI to consider a factory’s capacity and competencies, its past performance, results of all the inspections performed there, the types of defects found, and actions taken to eliminate those problems, among other factors. To see the results of an analysis, an operator uses a graphical interface or conducts a chat with the system.

These insights can help a company avoid costly production errors and protect a brand’s reputation. Also, by discovering which factories pose the greatest risk and which perform well, a company can save money on inspections.

“If there’s a low-risk factory that we’re sure will have good quality, the brand doesn’t need to have its own inspectors or pay a third party,” says Diego Pienknagura, Inspectorio’s executive vice president of strategy and business operations.

“Because Inspectorio is rooted in AI, it will be able to help us mitigate risk,” says Marienne Hill-Treadway, senior vice president, sourcing operations at Centric Brands, a New York-based apparel and accessories firm that contracted to implement three of Inspectorio’s solutions in 2024.

“It will give us a lot of data and KPIs (key performance indicators) that we didn’t have before,” she notes. This will help the company work more collaboratively with its vendors.

“Also—and this is a big one—we will be able to self-certify good performing vendors, so they can just report in to us, without having to spend money and time traveling to remote places to conduct an inspection,” Hill-Treadway adds.

One Dispatcher is a Bot

Route optimization was the first goal when Senpex, a last-mile logistics service based in San Jose, California, started to implement AI. More recently, Senpex has been training AI to work side by side with human staff.

“We keep five dispatchers, and one of them is a bot,” says Anar Mammadov, the company’s co-founder and CEO.

“That bot not only informs or helps with messages, but it also does actionable items, like canceling routes and giving them to someone else,” he says.

This kind of automation is critical when managing as many as 50,000 local couriers, especially during rush hour.

Senpex has also trained AI bots to hire new drivers and bring them onto its platform. When a driver uses Senpex’s app to apply to the company, a bot conducts an interview while verifying the person’s ID and physical location and doing a background check on social media. The software considers current demand for service across the United States, and how well Senpex is meeting that demand, to determine how many couriers to hire in various locations.

AI plays a supplemental role in sales at Senpex, too, retrieving information to help sales agents serve customers.

The company would like to use AI bots to acquire new customers and interact with existing ones. But right now, its bots aren’t sophisticated enough to offer information and services to meet the needs of demanding customers, based on data analytics. “When educating the AI, we need to explain not just how to do the sale, but how to do the sale based on the data,” Mammadov says.

You’ve Got Mail

AI-based applications draw much of their data from the digital systems that shippers and service partners use to run their organizations. But even in an age of web portals, apps, and APIs, many shippers still prefer unstructured forms of communication, especially email.

When a shipper sends an email to request a quote or ask when a shipment will arrive, someone on the provider’s staff must query a digital system, retrieve the results, and send an email reply.

To speed up those routine transactions, third-party logistics company C.H. Robinson has developed proprietary AI-based technology that reads incoming emails, extracts the pertinent information and fulfills the shipper’s request. It does this even when the email lacks some of the necessary details—for example, if the message says, “What will it cost to ship 20 pallets of widgets to our customer in Omaha?”

Like a human agent who has been serving the same account for years, the AI fills in data such as the customer’s address and the weight of the widgets.

“We use the AI to not only read the emails but also to get into the data, accessing all of it from our own systems,” says Megan Orth, senior director of commercial connectivity at C.H. Robinson. “We can tell the customer, ‘If you can’t invest in a TMS, that’s okay. We will give you the same experience.’”

As of October 2024, C.H. Robinson was using its new technology to respond to emailed pricing requests and load tenders. It’s also piloting the use of generative AI to interact with carriers that aren’t sending automated status updates.

The company reported in October that it was using AI to deliver 2,600 quotes daily, with a 32-second response time, and taking 90 seconds each to process 5,500 shipment orders daily.

Applying AI to these transactions makes them more accurate as well as faster. Automation also frees account agents to focus on high-value activities rather than routine tasks.

“Everybody’s learning every day how to apply AI where it makes sense,” Orth says. “AI is not a buzzword. It is real. It works.”

---

Broad vs. Narrow

The AI that excites most people these days is generative AI, the technology embodied by ChatGPT and similar systems. This AI can create new content—essays, pictures, answers to just about any question—based on the oceans of data on which it was trained.

Generative AI is not the best fit for many supply chain challenges, notes Keith Moore, CEO of AutoScheduler.AI in Austin, Texas. That’s because supply chains don’t need models trained on a world’s worth of data, but rather on data from a specific operation. The right tool for the job is what Moore calls narrow AI.

“In narrow AI, we take a singular task that can be performed by a person and do it more accurately or faster,” Moore says.

A large language model such as ChatGPT can’t tell you how to run a warehouse because it doesn’t have access to that facility’s data. But that type of AI, with its natural language abilities, can play certain roles in the supply chain. “We can train a generative AI model—almost a narrow generative AI model—on interpreting our results,” Moore says.

For example, in a warehouse that uses AutoScheduler’s platform, a user could say, in plain English, “I’m a site leader: tell me the 10 things I need to know about.” Based on that person’s role and the current operational data, the AI would respond with an update on inventory that’s running late, staffing needs throughout the current shift, and other concerns. “That’s where it’s extremely valuable,” Moore says.

---

Seven Questions to Ask AI Vendors

By Erica Frank, VP of Marketing, Optimal Dynamics

It can be overwhelming to receive emails and phone calls from artificial intelligence (AI) vendors who claim their technology is the solution to all your supply chain problems. But companies can find themselves deep into a big investment without knowing the exact problem the software will solve. These seven questions are the perfect starting point to ensure you end up with the right AI solution to solve your specific challenges.

1. Can you demonstrate how your platform solves my problems?

Once you’ve done the prep work internally to determine exactly what problem you need to solve and how technology can help, then it’s time to start exploring how AI-based software could be a solution.

A vendor should be able to demonstrate, using proven customer data, how it can solve your challenge and the potential results you can expect before the purchase process begins. This is also the right time to involve multiple members of your team from different departments as they might have different perspectives on the problem to solve. It’s helpful to see how a vendor tackles addressing the different perspectives within your company as well.

2. How does your platform help users in my industry?

If the vendor has no experience within the supply chain or transportation vertical, it should be an immediate red flag that it might not be the right fit. Without past examples of success, the amount of customization to make it work for you can become pricey fast and you don’t want to end up as the guinea pig either.

The vendor in question should have past case studies and success stories to share focused on companies within the supply chain, logistics, and transportation space. If possible, they also should have one for a company with a use case like yours, before moving forward. The vendor’s software should also be easily customizable to match your organization, network preferences, and any capacity restraints.

3. Can you explain how it works?

There is a difference between an AI technology vendor that has built an algorithm or a set of tools that analyzes data and one that uses AI to advance that data. Asking these “how” focused questions can help draw a clear connection between the application’s use of AI and the solution to your problem that launched your search for a vendor in the first place. Understanding “how” the algorithm works and why it’s providing certain recommendations will help foster trust and confidence in the tools.

Confirming concept and proof of value within your own data (see question #1), is essential. This step will confirm that the vendor is using the right tools to power a real application that delivers benefits, value, and solutions to its users and is more than just hype.

4. Is the solution integrated with our current tech stack?

This is a key question because the answer will also determine how fast a new system can be integrated and running. Connectivity is one of the highest priorities for tech buyers. End users don’t want to visit multiple interfaces and dashboards that interrupt their existing workflows, but rather look for ways to supercharge existing technology investments.

The technology should make the user’s lives easier, not harder. When a solution is easily integrated into a company’s current tech stack, the benefits of the new technology should be immediate without sacrificing efficiency.

5. Can your team help us embrace and manage change?

Even when a solution fits smoothly into your current tech stack, there will inevitably be a transition period. Team members must learn to use the technology, and incorporate the new tools, data, and information. Sometimes it may require a new way of thinking and carrying out day-to-day operations.

At this point in vendor selection, it’s important to talk to the vendor’s references to learn more about the implementation process and learn more about the ongoing support that they will provide during the ramp-up period as well as ongoing. Ask about post-deployment engagement levels, support, and overall satisfaction. Clarify customer support capabilities and dedicated teams and how they differ under various pricing plans. A vendor that provides ongoing support and invests in the success of the user’s organization is the mark of a true partner.

6. How well does the system manage customization and scalability?

The technology should be customizable to meet your organization’s specific needs and help it grow over time. Startups and smaller businesses need platforms that can be configured to help them solve the problems they have right now and into the future as they achieve success and scale. Even larger, more established businesses can always continue to optimize operations and uncover new growth opportunities.

Ask vendors how their solution can help support growth including the implications such as acquiring a new company or including more data points. In addition, ask about the vendor’s roadmap and path to growth, what new solutions they are looking to add in the future, and how those can support your company.

7. What can I start doing to prepare?

This isn’t necessarily a question to ask of a vendor, but rather of your internal team. The most successful implementations are the ones where the customer is aware of how the technology can transform the business and that starts at the executive level. Even if you aren’t ready to immediately implement a new solution, it’s still possible to plan for the future. Think about which data points are important, how you’re entering the data, how operators are putting information into comment fields, how communication is happening throughout the supply chain, and where operations can be transformed.

---

What’s Up Next?

The first wave of AI solutions for the supply chain focuses largely on helping humans do their work better, whether by supporting decision-making or automating routine tasks, says Katelyn Foley, president of UP.Labs.

For future applications, one of the next steps will be to integrate AI with hardware—for example, in an automated warehouse. “There’s promise in humanoid robots,” Foley says.

AI could also be applied to “digital twins,” creating digital models of supply chain networks and using them to run scenarios that can, for example, calculate the risk that a natural disaster or political upheaval will disrupt supply from a particular vendor.

“It would start out small: looking at this network or your full supply chain, what are the most critical vendors?” says Louis Matthews, venture CEO at UP.Labs. “And then expand it out to Tier 2 and Tier 3.”

Other possible uses for AI that UP.Labs plans to explore include cyber security and the physical security of freight in transit.

---