(FORTUNE Magazine) – Maybe it was the Cisco golf tee that inspired Chris Sinton's big idea. Or was it the Cisco coffee mug? Or the Cisco baseball cap? Anyway, it wasn't the routers and switches and other networking gear that are the San Jose company's main stock-in-trade. After all, Sinton's specialty was distributing marketing materials to customers, everything from technical brochures to trinkets, all of which, in typical Cisco fashion, customers had to pay for. Yet on the moth wings of that big idea--selling a gamut of goods and services to other businesses over the Internet, which Cisco does more of now than anybody else--the company has been able to protect its impressive margins, allowing it to underwrite a stream of new products. And the market is cheering: In three years Cisco's value has swelled from $15 billion to $100 billion today.

Sinton was an unlikely apostle for the Internet. By his own admission, the Southern Oregon University marketing dropout is one of the lowest-tech people at Cisco. "I don't talk in acronyms," he confesses. But he does speak the language of customers. He knew they hated wasting manpower phoning and faxing in orders, and ordering hats and mugs from printed, out-of-date price lists. Just as e.Schwab was getting started, Sinton began selling his golf tees and brochures on the Internet. And the Net was telling him something: Demand for souvenirs more than quintupled, and ten times as many people requested technical material as before.

Meanwhile, managers from other corners of Cisco were discovering their own ingenious ways of exploiting the medium. These experiments weren't mandated by top management. Typically, a department head would set up a skunkworks, asking a couple of engineers, working on a shoestring, to see how far a daring idea might take them.

By interesting marketing people two levels above him, Sinton was granted 15 minutes to make his pitch at a meeting of senior executives. He praised these other experiments and urged the company to turn them into a whole new way of doing business. Cisco could save a hell of a lot of money and serve customers far better, he asserted, by E-selling not only $10 promotional headgear and technical-support services but even its precious $1.5 million routers.

Sinton's presentation roused the room. E-commerce quickly became a corporate crusade. Today, Cisco--which sells 80% of all Internet hardware--has turned experiments like Sinton's into the backbone of its business. Since mid-1996 it has moved 57% of its sales--or around $1.3 billion this quarter--in routers, switches, and other gear onto the Web. Its goal for next year is 80%. Though the e.Schwabs and other business-to-consumer sites get the spotlight, companies last year sold other companies around $8 billion of everything from computers to airplane parts--four times the total sales to consumers. Forrester Research predicts sales will rise to $17 billion this year and will mushroom to $327 billion by 2000. Cisco, along with pioneers like Boeing and Dell, is proving that business-to-business selling is E-commerce's killer app.

Sinton may have been the first to see how it all fit together and grab top management's attention. But Cisco's first Internet breakthrough--without which massive E-selling couldn't have happened--came in its Technical Assistance Center, which provides after-sales service. Networking gear is so complex that after a sale, customers are in continual contact with their supplier. Knowing that, they buy products as much for the quality of after-sales service as for the items' price or features.

By 1994 the Technical Assistance Center was facing a staffing crisis. Service engineers are highly trained and extremely scarce, yet at Cisco they spent most of their day fielding routine questions about minor malfunctions, leaving not enough time to deal with the really tough technical challenges. Customers even called them to order software. This created a company-strangling constraint: If Cisco couldn't hire new engineers fast enough or free up the ones it had, it would have to cut back sales of routers and switches.

The solution, thought Brad Wright, the center's head, was to automate all the routine stuff on the Internet and let buyers serve themselves. With the backing of Doug Allred, the head of all sales and support services, Wright started his own skunkworks, charging several of his engineers with developing programs that could answer queries online. The system would translate a network engineer's fuzzy inquiry ("Cannot connect to remote server'') into a standard description of a familiar problem; then it would provide, say, the four most likely explanations onscreen, allowing the engineer to avoid a wilderness of blind alleys. Within 90 days Wright's team had put the most frequent questions on the Web. It also set up a program that let customers choose and download software.

The reaction was astounding. Weary of playing phone tag with busy engineers from nine to five, customers flocked to the Internet for effective 24-hour service. Calls and faxes dwindled. Cisco's sales have jumped fourfold since 1995, while its engineering support staff has merely doubled, to 800. Without automated sales support, Cisco estimates it would need well over 1,000 additional engineers. Estimated savings: $75 million a year, plus another $250 million it keeps by distributing support software over the Internet rather than transferring it to disks and mailing them to customers. Not to mention the billions in sales the company might have forgone if it hadn't been able to find those 1,000 extra engineers.

In a different corner of the company, another adventurous manager was at work. In early 1995, Linda Thom Rosiak, the new head of customer service, was fretting over Cisco's ordering process. The "dissatisfier," as Rosiak puts it, was the time and manpower it took to push orders from Cisco's customers to its plants or suppliers.

The cause of the delays was the multitude of errors in the orders, which invariably arrived by fax. All of Cisco's products are custom-built. Each has about a dozen major elements, including memory, power supply, software, and cables. Cisco offers dozens of choices for each, but many of the combinations don't work together. For example, customers frequently chose insufficient memory to handle their choice of software. Making matters worse, customers found the prices for 13,000 parts in often out-of-date catalogs the thickness of phone books. When Cisco received orders with the wrong prices or configurations, which was 40% of the time, it simply faxed them back to the customers.

"It was a disaster," says Frank Santafemia, head of network installations for Sprint, which, like most of Cisco's customers, schedules delivery of the products that go into the giant networking systems that it builds and operates for its corporate clients. "We'd be faxing incorrect purchase orders back and forth. Meanwhile, we'd need that router to build the project. Those errors could delay jobs for weeks."

Putting the sales process on the Internet, Rosiak realized, would eliminate screwups. Customers could complete their projects much faster, and Cisco would avoid the cost of hiring an avalanche of new people to catch the errors.

So Rosiak helped start yet more skunkworks. A key collaborator was Cisco's chief information officer, Peter Solvik. "If we could leverage the Internet in technical support, I knew we could use it for every step in selling and servicing our products," he says. Solvik assigned Rosiak a couple of his own engineers to help design programs linking customers to Cisco. The first one, called Status Agent, let customers track the progress of their orders online. The next involved posting the prices of all Cisco products. The third was the most momentous of all--but that one was going to require the support of top management.

Around this time, a venture capital firm had introduced Solvik to Calico Technology, a tiny startup. Calico made software that enabled customers to unerringly select compatible parts not just for simple machines like PCs but also for big-ticket items like routers. At a conference in Monaco, Solvik asked CEO John Chambers for a few million dollars to buy Calico. He told Chambers he could make the key ingredient needed for Internet selling--online configuration of Cisco products--customer-friendly. Chambers, by now an E-commerce zealot, said yes.

Cisco offered routers and switches for the first time over its new business-to-business Website, Cisco Connection Online, two years ago. Customers quickly saw the advantage of picking out prices and configuring products electronically. They simply click onto a program called Configuration Agent, which walks them through the dozen major components that go into a router. If they choose the wrong combination of circuitboards, for example, the program posts an error message and guides them to an acceptable choice. Once the right item is selected, its current price pops up automatically.

Resellers are flocking to CCO because it lets them get their equipment and finish their jobs much more quickly. At Sprint it used to take 60 days from the signing of a contract to complete a networking project. Now, thanks partly to the efficiency of ordering Cisco equipment online, it takes 35 to 45 days. Sprint has also been able to cut its order-processing staff from 21 to six, allowing the other 15 employees to work instead on installing networks, a business that has doubled at Sprint since 1996.

Like its customers, Cisco is saving tons on order-processing workers. Rosiak has just 300 service agents handling all customer accounts; she says she would need 900 without the help of CCO. The difference represents another $20 million or so a year.

Is the Net bringing Cisco new business or just handling orders the company would have received anyway? Cisco's sales are growing by 30% a year, far more quickly than those of rivals like 3Com and Bay Networks. Resellers who used to spread their orders around are buying mainly from Cisco, thanks to the convenience and speed of CCO. Confirms Sprint's Santafemia: "We are doing more and more projects where Cisco is supplying all the equipment. It's just faster that way."

But E-commerce's biggest impact is on profits. At most high-tech companies, profitability shrinks as sales take off. That's because prices fall as new players enter the market. Though Cisco has not escaped that iron law, its Net operations have kept its selling and servicing costs--from personnel to printing to software distribution--far lower than its competitors'. All told, including savings harvested from manufacturing, the Internet is saving Cisco about $360 million a year and helping preserve its 32% operating margins.

Chris Sinton, the man who blazed the trail, is now a member of top management, overseeing the operation and design of CCO, his brainchild. He reflects: "I just knew the Net could be our business, that it could be a portal to our company." And to the history books, one might say. The visionary white paper he wrote in 1995 on the future of E-commerce is now preserved in the Smithsonian Institution archives. Not bad for a former souvenir salesman.