If Efficiency is No Picnic, Follow the Ant Trail…

Organized and extremely efficient, anthills have a thing or two to teach us about running a company. That's why experts are using ant colony rules to improve business operations:

If companies functioned like ant colonies, their processes would be much more efficient and in sync. Indeed, we have much to learn from ants, whose team efforts are expertly coordinated. For example, each anthill member performs a particular task and leaves a pheromone trail that informs its peers where it's been, what dangers to dodge, and which route will take them home the quickest. Why such a level of organization eludes us has fascinated Eric Bonabeau—a French scientist and believer of chaos theory, a branch of complexity science—for years. He has concluded that anthills run so smoothly because they don't depend on centralized control, and what's more, he says, companies can solve problems by studying these hyper-efficient ant colonies. In his 1999 book, Swarm Intelligence, he likened a company's many individual parts to "ants" and postulated that by focusing on these distinct entities, companies can find answers to questions that stump ordinary top-down analysis. For example, we can ascertain how one single delayed package can disrupt an entire supply chain or why expanding a highway can actually exacerbate traffic. And now, Bonabeau is taking his ideas from paper to practice—through consultancies that help companies with "agent-based modeling" techniques. Such consultancies, including Santa Fe's Biosgroup, where Bonabeau got his start, and Icosystem, which Bonabeau founded, produce algorithms that create extremely realistic simulations of all of an operation's functioning parts. The firms factor in virtual workers, products and customers, tell them to do different things, and observe what ensues. In fact, agent-based modeling is already helping several companies improve their supply chains and shorten time to market. Even the U.S. Office of Naval Research is using this ant-inspired method to build better unmanned aerial vehicles (UAVs). One company that is taking tips from ant colonies is Air Liquide, a French industrial gas giant, which has a very complicated supply chain. Delivering liquid oxygen, nitrogen, and other gases to 10,000 customers from over 300 sources through 30 depots with 200 trucks and 200 trailers, its supply chain can generate 3 trillion daily combinations from all its moving parts. Such a massive operation required 22 full-time logistics analysts to labor for nearly half a day just to come up with a delivery schedule. Biosgroup engineers lessened the swarm of logistics professionals needed by generating and tweaking computer simulations until they discovered the most effective combination of rules. Now, only one Air Liquide analyst sets the daily shipping and production schedules—and does so in roughly two hours. "There's no way a human could look at all the possibilities out there," says Air Liquide senior project manager Clarke Hayes. "So we turn the ants loose." Ants have also left a trail on Southwest Airlines' operations. Biosgroup simulated the different parts of Southwest's cargo shipping business—its airplane fleet, destinations, cargo, ground personnel, etc. The simulations backed up a theory long held by Southwest logistics experts—that the shortest route is not always the most efficient, what's important is the number of hands handling the cargo. According to Southwest, this insight saves the company $2 million annually in labor costs. Agent-based modeling is not only about ants either. "Inspiration comes from ants, but we'll use any rules to help solve problems," says Bonabeau, who founded Massachusetts-based Icosystem after leaving Biosgroup in 2000. Currently, Icosystem is applying an evolved form of agent-based modeling to help Indianapolis-based pharmaceutical company, Ely Lilly, bring drugs to market quicker. By plugging in factors such as people, drugs, regulatory requirements, and other parts, Icosystem recently created a model of the company's drug development processes. By pinpointing and implementing the correct rules, Bonabeau aims to quicken Lilly's development time by as much as 80%. In addition, the consultancy is aiding insurance giant Humana, by generating models that will shed some light on how people select their own health plans. Icosystem also counts the U.S. Office of Naval Research as one of its clients. The consulting firm is helping the organization improve the design of its UAVs. Such unattended vehicles often perform poorly as a team—failing to react to threats and amassing in certain areas while neglecting others—because they are directed by a centralized ground command. To make these vehicles smarter, Bonabeau and coworkers have produced simulations in which virtual drones are programmed to adhere to hundreds of rules—keep a certain distance away from other drones, cover areas not patrolled by other drones, etc. The initial objective is to let drones talk to each other, and by 2020, to have armies of such vehicles communicate and execute coordinated attacks. But applying anthill lessons to organizations is sometimes met with some resistance. "Managers don't like to give up control," says Bonabeau. There is also the issue of cost. Simulation games demand math and programming expertise as well as extensive business data in order to be useful. Clients spend as much as $200,000 a month for such services. However, these drawbacks have not deterred the growing number of companies that are deriving lessons from ant colonies. IBM recently started working on its own agent-based modeling programs for e-commerce software. Even the European Union is recognizing the wisdom of ant colonies, financing a three-year modeling research endeavor at the Santa Fe Institute. Bonabeau says that eventually he hopes the technology will fix problems before they even occur. But a more immediate goal is generating big profits for his clients. "We need a company to make $1 billion from this technology," he says. Sources: The Wisdom of the Anthill
Thomas Mucha
Business 2.0, Nov. 2002

All Topics