In a summer when "The Matrix: Reloaded" reigns at the box office, you probably won't be surprised to know that computers are already making decisions about our lives without any human intervention. Artificial intelligence has become a mundane reality, used in Web services such as Amazon.com's, and to control production lines, city traffic patterns, telephone call routing and even some banking functions. But the logistics and transportation sectors have so far been reluctant to implement so-called smart software, for reasons of money, time and plain old fear.
All that is about to change, according to several experts in logistics technology. "Over the next nine to 12 months you'll see significant pilot projects taking place, at which time the concept will either be proven or disproven," says John Karonis, director of fulfillment technology at Kurt Salmon Associates in Princeton, N.J. "We're confident that it will prove to be a worthwhile endeavor and that we'll then see it rolled out on a much larger scale." Karonis has been working on a project to combine the power of radio-frequency identification (RFID) tags with intelligent software in a way that allows a computer to decide how to fix problems without human intervention every time there's a glitch in the movement of goods.
But what is intelligent software? Dr. Noel Greis, director of The Center for Logistics and Digital Strategy at The Kenan Center, University of North Carolina-Chapel Hill , explains that it's a type of artificial intelligence (AI). AI falls into two broad categories, she says. One is aligned with robotics and artificial vision, the sort of science that holds the promise of an electronic butler who hands you a drink and makes dinner when you get home, or an order-picking machine that would notice if a product was damaged and do something about it. But the other side includes what's known as intelligent software "agents." Also known as "bots," these are software packets that act as autonomous, decision-making entities, capable of coming up with solutions to problems and acting on them automatically.
Intelligent agents can be very simple. A good example is the way Amazon.com offers you a list of books you might like to buy in addition to the one you've just chosen. That's simply an agent that's programmed to think: "If this person orders this book , then I will automatically offer him or her these books, based on choices by other people who ordered the same book ." A more sophisticated agent will keep your personal history of books ordered and suggest new publications that fall within your recorded fields of interest when they become available, also a current feature on Amazon.com. It's only a matter of time, agree Greis and other academics and consultants, before intelligent agents get put to work in the logistics and warehousing industries.
How would they be useful?
First of all, intelligent agents cut out the delays associated with waiting for a human reaction to a glitch in cargo movement. Telecommunications companies such as British Telecom in the U.K. use intelligent software to automatically route calls through the cheapest and most readily available lines. The same could be done with trucks navigating congested roads, or packages moving through a distribution center. Another application that surfaced in the crazy days of the transport dot-com boom was the automated negotiation of spot-market transportation buying. This typically involves fast-paced juggling of rates and availability measured against the performance records of known and unknown carriers. Software that compares apples to apples in the blink of an eye, then accepts or rejects bids could be highly useful. It didn't catch on in a public online auction scenario, but it could work in a private one.
However, Karonis of Kurt Salmon says it's when you combine intelligent software with other technologies—particularly data-collection devices —that things really get exciting. That's because software that makes decisions in real time needs better and more accurate data than is commonly available along the supply chain.
"RFID means more accurate and timely data, but if I don't have a decision engine to do something with that data and I'm just forcing it into the old processes, I'm not going to be able to do anything useful with that data," says Karonis. "By the same token I could deploy intelligent agents to make more intelligent and timely decisions, but if I'm using old data, the value of those decisions is going to be questionable. It's when you put them together you have more accurate, timely data leading to more accurate, timely decisions and that's where the real benefit lies."
Combining quick logistics management decisions with real-time data is the way forward for warehousing and supply chain expertise, says Greg Schlegel, former president of APICS -The Educational Society for Resource Management and a senior manager in IBM's ERP/Supply Chain Management Group. Schlegel predict s wide spread deployment of intelligent software to help that happen. "You're getting into neural networks where software can learn and make its own decisions and build learning trees about what to do and what not to do. From there, you get into predictive analysis, the ability to [resolve] problems before they arise. That's the kind of application that logistics and transportation managers are going to deploy."
So far, most of the work on getting logistics software to act intelligently is being done on university campuses. The Massachusetts Institute of Technology in Cambridge, the Robotics Institute at Carnegie Mellon University in Pittsburgh, The Center for Logistics and Digital Strategy at the University of North Carolina-Chapel Hill, and the Department of Computer Science and Engineering at the University of Minnesota have all been working on intelligent logistics software in one form or another. In fact, they all have pilot projects under way in the commercial world, but most of the test subjects prefer to remain silent on early adoption. "They're not normally discussing it because they consider it a competitive advantage to be more cost-effective and efficient," says Schlegel. The truth is that adoption rates are low, so far. "There's probably more hype than actual adoption out there right now," says Dr. Steve Smith,a colleague of Dr. Greis's at UNC.
One of the barriers to adoption is agreeing on data exchange standards, says Karl Waldman, president of software vendor OAT Systems in Wa tertown, Mass. In conjunction with MIT's Auto-ID Center, OAT is working with Gillette to take information gathered via RFID tags in retail outlets and feeding that back into the company's warehouse management and replenishment systems. Up-to-the-minute stocking data isn't worth much if it's in a language the replenishment system can't understand. "Standards are a big problem," says Waldman. "CIOs are looking for something standardized so they don't have to integrate it all later." The Auto-ID Center is a joint industry/MIT initiative to help establish and promote those standards, and OAT has developed a data handling framework called Savant that can be integrated into existing systems to foster standardized data exchange.
But problems with the human element also provide a barrier, Waldman says. "A major [obstacle] is education. Everybody 's been using ERP (enterprise resource planning) and WMS (warehouse management systems) for a number of years, and those systems all represent inventory in a very simple way, so there's a lack of understanding about the types of visibility you can get with RFID and auto ID. We have to spend a lot of time educating people. When they understand there's a whole lot more stuff they can do, their eyes light up."
Bytes and pieces
Most companies are still learning how to use logistics management software that falls below the definition of intelligent. Exception alerts are a good example. These will monitor the flow of goods through a warehouse or supply chain and send out automatic alerts when something goes wrong, prompting a management decision from a human being. For example, Optum is helping Lucent Technologies coordinate complex production and delivery functions. "Their whole goal is to get around 80 suppliers for any given order to ship so that the order all comes together in a three-day window for delivery to a job site," explains John Davies, cofounder and vice president of product marketing at Optum, based in White Plains, N.Y. "If one of the key suppliers producing a critical component can't ship it on time, they provide a message to us and we will automatically route messages to all the other suppliers that the date is going to have to be pushed back."
At a high level of automation,this would constitute intelligent software. But, in this case, the software isn't allowed to decide on a new delivery date without consulting a human manager. "We'll send out a new date but we want someone to say: 'Yes, that's the right date,'" Davies says. He says programming intelligent agents to make reliably good decisions according to the myriad possible situations that may occur in a complex supply chain is currently too much effort for too little return."There are too many variables; it's too hard to write the rules," Davies says. "Humans are still good to have involved in the supply chain."
Davies and others agree that there's reluctance among logistics managers to hand over responsibility for crucial decisions to the machines. Optum's software does help automate some order fulfillment decisions for InvaCare, a maker of medical equipment,making last-minute decisions about how to fill orders based on real-time information about what's rolling off the production line and how demand has changed. "But that's a point solution. It's not like two agents getting together and negotiating and going off automatically," Davies says. "InvaCare wouldn't want those agents to expand into ordering supplier materials on the basis of that information."
Point solutions—or fitting an intelligent agent to a single business function such as cross docking—represent an ideal way to start with intelligent software, says Greis. "These are bottom-up technologies. You identify a problem and then develop an application to support it," she says."It's not like installing a huge SAP system. It's more about pulling out a particular part of the operation and having the agents work on it." Greis says this can be cheap compared to putting in a huge mainframe system. "The applications that we've done are designed to be overlays on existing systems and as inexpensive as you need to have them be," she reports.
IBM's Schlegel says logistics is simply taking time to catch up with other industries that are already exploring the benefits of intelligent software. "Artificial intelligence is being [used] in a big way in banks and financial institutions. They were the first to use neural networks and network systems," Schlegel says. He says banks have a lot to gain from automating computer operations and taking out "touch points" where a human has to enter information, since their business is mostly about data processing and protocols. After the financial services industry, manufacturing became the second group to adopt intelligent software. "They're star ting to embrace the use of message alerts for their supply chains internally," says Schlegel. "Now, the third industry is logistics. They're not embracing it yet, but they're talking about how to leverage it."
"Any time you have complexity in a business process, you can use agents to support a human's decision-making capability," says Greis. "Whether it's logistics or warehousing, it's about figuring out what decisions people have to make and asking whether an agent can make that decision better, faster or in a more cost-effective way."