It's funny what can happen over a turkey sandwich. Witness what has come out of the kitchen at Xerox Corp.'s Palo Alto Research Center (PARC). Those familiar with PARC know it for its sharp minds and creative inventions. But locally it's known for something else: a cafeteria that boasts good, cheap food and a breathtaking view of the valleys surrounding San Francisco. Because these treasures are open to anyone with a Xerox badge, the technicians who service Xerox equipment in the area often stop by PARC for lunch. So it was here that they-the people who spend their working hours digging deep into the guts of complex machinery-met some of the top research minds who spend their days wrapping their brains around concepts that hover between the brilliant and the preposterous. As a result of those meetings, researchers and Xerox technicians came together to develop a system for managing knowledge that flourished as a grassroots effort in a company that's not known for paying attention to the grass.
Now Xerox technicians are using knowledge management to share how they fix machines better and more naturally than most companies dream about. Call it an accident-a collision of the real world and the cerebral world that resulted in something thousands of Xerox employees use every day. Xerox calls it financial good fortune-the system that eventually came out of those PARC lunches has saved the company millions. It also carried a critical lesson for anyone trying to make sense of the bewildering world of knowledge management: True knowledge sharing has to start and end with the people who have the knowledge. This, then, is the story of how KM visited Xerox and left it a better place.
Tucked away behind one of Silicon Valley's main thoroughfares, PARC is funded mostly by Xerox and partly by the government. From behind its walls emerged such inventions as the first commercial mouse, laser printing and flat panel displays, while hundreds of other projects remain sequestered.
Our story begins with researchers working on artificial intelligence who wanted to see if they could replace the paper documentation that Xerox technicians used on the road with an electronic form. Some thought their lunchtime contact with technicians might help them test their theories of artificial intelligence (AI). They thought, "What if we built an expert system-a system that can "think" the way a person does-that would do the job of the documentation, suggesting fixes based on symptoms?"
The team-led by researcher Danny Bobrow and including software engineer Bob Cheslow and researchers Mark Shirley and David Bell-found that it was indeed possible to build software that could do just that. But when they showed their first efforts to technicians, the response was underwhelming.
What kept technicians from finding fixes was not that the documentation was paper-based but that it didn't address all the potential problems. And not all problems were predictable. Machines in certain regions could react to extreme temperatures in different ways. A can of Mountain Dew overturned in one part of a machine could wreak havoc in another seemingly unconnected part. Technicians could handle these mishaps quickly only if they had seen them before or if another technician had run into a similar problem and shared the results.
Once the conversations with technicians revealed this gap in information sharing, the researchers realized that AI was the wrong approach. What Xerox needed instead was knowledge management. "It wasn't a smart computer program that was going to fix these things," says Cheslow. "It was sharing the best ways to make these repairs."
On The Road Again
Ever bring your car to the mechanic because of a strange rattle only to have it fall mute as you pull into the shop? It happens with copiers too. Technicians call these problems intermittent, and they can be the most time-consuming and frustrating. More to the point, they rarely get documented.
When a Xerox service technician faces a broken machine, it's man against beast, Captain Ahab and the great white whale. Typically, technicians service between two and five machines per day, and customers eager to make copies don't look kindly on a visit that doesn't result in a fix. What's more, if a technician can't fix a machine, Xerox's policy is to replace it-sometimes to the tune of tens of thousands of dollars.
When the PARC researchers realized they needed to look at the way technicians work, they spent time in the field, following the technicians from call to call. What they observed proved invaluable-knowledge sharing was already unofficially ingrained in the organization. Most striking was not how technicians solved common problems, but what they did when they came up against a tricky, intermittent one. Often they called one another on radios provided by the company. And in informal gatherings they shared vexing problems and their fixes.
What the researchers observed on the road matched what their efforts creating an AI system had revealed. "Finally we understood that what they needed most was the kind of help they get from each other," say Bobrow. And that was the beginning of something small that turned into something really big.
Meanwhile, another PARC researcher-a Frenchman named Olivier Raiman-was busy comparing the way French and U.S. technicians worked. Traveling back to France, he discovered that, while French technicians appeared to work from immaculate, uniform documentation put out by headquarters, their real solutions also came from a second set of documentation-notes they carried with them detailing what they'd learned. Raiman set up a series of workshops with expert technicians to ask what information they wanted to see shared. The workshops revealed that they had all come across similar problems and were eager to hear how others had fixed them. So Raiman started collecting the tips technicians had discovered and put them in a central database. It was from that database that, back in Palo Alto, researchers started building the first laptop-based knowledge-sharing system.
PARC researchers took the first iteration to France and began a series of exhaustive sessions with the French experts in the Xerox headquarters outside Paris. In those sessions, something magical happened: The system took on the shape of the people working on it, evolving with each suggestion from the actual users.
But the worldwide customer service group didn't take the project seriously. "Nobody believed that the knowledge of the technicians was really valuable," says Bobrow. So, working stealthily, outside the realm of worldwide management, Bobrow's team gave laptops and the fledgling program to 40 technicians and matched them with a control group of technicians who relied solely on their own knowledge when fixing machines. After two months, the group with the laptops had 10 percent lower costs and 10 percent lower service time than those without-and the control group was jealous of those with the system.
These results were enough to convince PARC to front the money to build a new system, called Eureka, which was initially deployed over the French Minitel (a dial-up service from the French telephone company). As the workshops to improve Eureka continued, the researchers and technicians addressed knowledge-sharing issues: How would technicians know whether a tip was any good? Would they get credit? How would they know their tip wouldn't disappear into a black hole? From these conversations emerged the validation system, in which experts in a particular area would test submitted tips before sending them live. Useless tips were discarded; others were certified as valuable or edited as necessary. It was agreed that tips would be validated within a few days, and the submitter's name would appear alongside the tip as both reward and incentive.
Management Says, "No, Thanks"
Even with the obvious excitement Eureka was generating in France, the project came up against a fairly formidable speed bump. The top dogs at Xerox remained unconvinced that such knowledge sharing was worth its salt. Accustomed to ensuring quality of service through uniform manuals, management was reluctant to let service technicians have control over machine fixes. "The people at the top felt that technicians were not terribly clever and they didn't do very much; they were mostly just following the instructions that came out of the manuals," says Bobrow.
PARC software engineer Cheslow agrees. "The really hard-core resistance was that we can't have service reps cooking up these crazy ideas." And management worried that costs would rise if the tip system spread. But the opposite was true. In the first year of Eureka's existence in France, the system was seeing at least one new tip per day, with more than 20 percent of technicians having had submitted at least one tip and almost all using the system at least twice a week. And costs were falling, not rising.
Buoyed by the positive reaction in France, PARC wanted to press on. Having seen how eager French customer service operations for Xerox in Wisconsin and Iowa, his group participated in the U.S. technicians were to share knowledge-and also eager to help create the system that would let them share it-they believed they had discovered something incredible: Listening to users was by far the best way to create a knowledge-sharing system. That became the heart of Eureka, and its appeal was pure and simple. "This was not something that was cooked up in the big brains at PARC and given as a generous gift to the service community," says Jack Whalen, a behavioral scientist who joined the Eureka team three years ago to lend his expertise to figure out how to roll out the system in the United States. "The ideas and practices we were trying to enhance were already there. That's why it works," says Whalen. And so it grew.
Next the researchers turned their attention toward Canada, which had about 1,200 technicians-the same number as France-and had just completed a laptop rollout. Still unsanctioned by management, Eureka was distributed quietly at first, handed from user to user on a floppy disk, like a bootleg Springsteen tape.
"We just did it. It was a guerrilla sort of thing," says Cheslow. "And it was clear this added to the sense of ownership." Then, with the results so good in France and the anecdotal evidence increasing in Canada, management started to pay attention, opening its wallet and doling out money and support bit by bit. By mid-1996, management had begun to get behind it, and Eureka became official in Canada. That brought about new challenges, like reconciling the community-driven approach that had been the soul of Eureka with management's expectations for a top-down approach. In addition, some technicians expressed skepticism about the new technology. To quiet those concerns, one of the Canadian experts became an evangelist for the system, traveling around the country to demonstrate the glories of the new system. And the Eureka team remained adamant that the community be involved in suggesting and implementing change.
Then came the big one. "Eureka was planted in Canada based on seeds that were already there. But as soon as it started getting planted, there were other places with connections one way or another that were willing to go find it and get it," says Cheslow. One of those places was the United States, where the sheer size of the technician force-10,000-made it impossible to use the true guerrilla approach that had worked in the other two countries. Even so, it began to spread like a giant ink stain until management agreed to send it through official distribution channels.
Craig Matzke remembers some resistance as the project reached the United States. As the manager of rollout early on. Accustomed to things being handed down from the top without much regard for actual work practices, technicians were reluctant to cheer when they heard another project was on the way. "They thought, Here's another program that's going to create more work,'" Matzke says. But as word about Eureka spread from people who had actually tried it and liked it, the idea caught on.
By 1998, Eureka was officially deployed in the United States and began to make its way around the globe. Today the system has more than 15,000 user tips, with more being added every day. The hope is that by 2002 it will be distributed worldwide to the company's 25,000 technicians. And already success stories abound. One technician in Montreal authored a tip about a 50-cent fuse-holder replacement that caused a chronic problem with a high-speed color copier. A Brazilian technician had the same problem, and his customer wanted the $40,000 machine replaced. When he found the tip from Montreal in Eureka, he fixed the machine in minutes. Current estimates have Eureka saving Xerox at least $7 million in time and replacement costs. It's tales like those that make senior management happy.
But beyond the knock-'em-dead stories, technicians like Chuck Rutkowski, a senior customer service engineer at Xerox in Madison, Wis., and a 28-year veteran of the company, use Eureka every day. Why? Because the information comes from the people who actually use it, so it is eminently useful. "If I want to know what temperature the water is, I'm going to ask the guys splashing around in the water," he says. "Not the guy standing on shore."
Busy technicians are starting to realize that taking the time to write a tip might pay off for them one day. "The more it works, the more likely you are to want to put something in," says Rutkowski. And then there's the motivation of having one's name in lights: "The fact is your name is there in front of thousands of your peers all over the world," says Cheslow.
But perhaps the best proof of Eureka's success is the way it has moved from a "last resort" tool to one of the first things technicians turn to.
Of course with growth comes growing pains. Now that Eureka is in its second iteration, the new and customer service operations for Xerox in Wisconsin and Iowa, his group participated in the U.S. improved Eureka II, it's no longer a tiny grassroots effort championed by a few and ignored by many. And that means new challenges, like balancing the system's purpose with its potential. Cheslow notes that people want Eureka to be a catchall for all corporate information, and while that's technically feasible it might not be wise. "We've been struggling with the tensions of building it into something that is 'the one' and holding true to the principles that this is for supporting the work practices," says Cheslow. And now that PARC researchers have moved on to other things-their job, after all, is to create, not to manage what they've created-it's up to Xerox to take Eureka to the next level.
Still, the lessons remain. "The breakthroughs here were in the way in which it was done, which was really tightly tied to what people's work was actually like," says behavioral scientist Whalen. "The heart and soul is so different from the way most corporations think about technical support."
So, in the end, did a little system started by a few researchers at PARC really change the way a whole organization looks at knowledge? Maybe. "I think there's a lot more recognition of the value of that frontline knowledge," says Whalen.
Did it change Xerox's acceptance of the importance of peer-to-peer knowledge sharing? "Absolutely," says Cheslow. "But it's hard to know yet if the lesson of working with the users has really been learned." Recently, the copier king has started on the path toward dethronement. The company posted third-quarter losses in October, and its stock has dipped into the single digits from a high of more than $60 in May 1999. Many Silicon Valley insiders intimate that part of Xerox's troubles spring from its failure to take advantage of ideas cooked up in PARC's labs. Now Xerox is seeking noncompetitive partners for the research center to help turn inventions into business plans and ideas into hard cash. But that doesn't mean there isn't a lot to be learned from the PARC's journey toward knowledge sharing. No matter how you slice it, it's funny what can happen over a turkey sandwich.