A collaboration between IBM Global Services and Research is yielding tools to keep the new online economy humming.
For the past seven years, while the world has been transformed by the Internet revolution, a second, related revolution has been quietly taking place within IBM. From virtually nothing, IBM Global Services -- a division formed in 1993 -- has grown into IBM's main source of revenue growth and the world's largest provider of information-technology (IT) services.
By almost any measure, Global Services has become one of the most important sectors of IBM's business. "From a people point of view, it is the largest part of the IBM company," says Carol Kovac, vice president of services, applications and solutions at IBM Research. "It is the fastest-growing part of the IBM company, and it is in the sweet spot of the IT marketplace right now." In 1998, Global Services employed 145,000 people worldwide and generated $30 billion in sales.
The engine driving this rapid growth is the rise of e-business. In the past, businesses were able to service most of their information technology needs in-house. But in the world of e-business, computing has moved from the back room to the front office; it is the heart of a business, pumping data to and from customers, partners and suppliers, who typically are each running different programs on different hardware. "The growing complexity of these systems means that, unless you are continually retraining and rehiring your information-services staff, you eventually will no longer have the skills you need to keep your enterprise alive," says Colin G. Harrison, director of global services research at the Thomas J. Watson Research Center, who is spearheading a partnership between IBM Research and Global Services. Increasingly, companies have decided to recover the capital they have sunk into in-house information services by outsourcing. "Companies are looking for a 'chauffeur' to navigate the complexity of incr
easingly sophisticated IT systems," Harrison explains.
"It's too much to expect that there's going to be any magic piece of software, or the perfect single platform, to make this transformation easily," says Kovac. "That's where IT services comes in, and it's why the largest part of the IT marketplace worldwide is services, and probably will be for the next decade and beyond."
To keep up with this growth and maintain its lead in the IT and business services industry, IBM Global Services is enlisting the help of IBM Research. "All IT services have been about labor," says Kovac. "While the price of hardware and software continues to drop, the price of labor increases." And as information technology grows more complex, companies need ever-more-skilled workers in ever-greater numbers, which increases the cost of obtaining those skills. "So Research's challenge," says Kovac, "is to look at ways that technology can either reinvent parts of the business we do today or invent entirely new businesses that make use of technology, especially to boost the productivity of labor."
Over the past two or three years, IBM Research has been exploring ways to apply its vast store of intellectual capital to the challenges facing Global Services. The developing relationship between Research and Global Services has already begun to yield dividends on three fronts, characterized by increasing time scales, according to IBM Global Services strategist Ellen Dulberger. First, she says, Research has helped "put out fires." For example, Research has enabled Global Services to optimize the allocation of resources to help customers recover from serious disruptions to business, such as the hurricanes that hit the U.S. East Coast last September. Research has also provided software to model scenarios of business disruptions caused by such disasters and has a scheduling engine, a tool that automatically optimizes the scheduling of customers' tests of the recovery systems.
In the second time scale -- up to 18 months out -- Research is helping Global Services improve efficiency and cut costs. An example is Research's work to boost the effectiveness of IT help desks. Finally, beyond the 18-month horizon, Research is "collaborating with us to anticipate market trends and their effects on customer requirements, as well as to develop and commercialize new technologies that might enable us to offer new services," says Dulberger. Those efforts mean tangible results for customers on a number of fronts.
The automated help desk
To the many companies that have chosen Global Services to maintain their information technology systems, IBM is a largely invisible presence. That is, until a router refuses to route or a server simply will not serve. Then, says George Welyczkowsky, director of Global Services' Service Delivery Technology Center, "the image of IBM Global Services is our call center." It follows that improving the call center experience -- making it faster, easier and more intelligent -- directly enhances customers' image of Global Services and their overall satisfaction.
To Welyczkowsky and his colleagues, who develop services for IBM's outsourcing customers, it was obvious that technology could provide the necessary improvements. "Many of the things that the call center can do," says Welyczkowsky, "we can make available via the Web. And that's what we call digital service delivery." To help realize the vision of digital service delivery, the Service Delivery Technology Center formed a partnership with Research to develop intelligent, Web-based help systems.
Another Global Services partnership with Research grew out of a similar need. Product Support Services (PSS) helps customers who own and operate their own systems. Some years ago, PSS began an initiative to streamline this process, reduce costs and use the Internet as the delivery vehicle. As the initiative progressed, it became obvious that the Internet presented an opportunity for new service capabilities and functions. It also presented a new set of challenges. Although rich in skills, Global Services recognized early on that an even broader range of skills and knowledge would be required to develop many of the desired new functions. Again, it was a natural step to team up with Research.
The first application that Welyczkowsky's team worked on with Research, code-named HelpNow, is an automated help system. Like many of the collaborations between Research and Global Services, HelpNow is an almost equal mix of technologies requested by Global Services and concepts that Research was already pursuing.
For example, the IT systems of most large companies are maintained by a patchwork of different vendors: there's one that maintains the LAN closets, another that maintains the networks, and so on. Communicating with these vendors means dealing with a Babel of different computer systems. Global Services approached Research with the idea of developing a system using XML, a powerful new approach to representing data on the Web, to allow seamless communication. "With XML we had a specific problem that we came to Research to solve," says Welyczkowsky. Such requests can have unexpected benefits. After finishing the project, Welyczkowsky recalls, "the researchers came to us and said, 'Hey, working on this has given us ideas for six or seven additional projects.'"
At the other extreme -- one where the basic problem had already been tackled -- was a knowledge management technology that could take a stack of text documents and automatically sort them into categories like a good librarian. The scientists at the Almaden Research Center who had developed this "e-classifier" technology saw that it had a direct application to the help desk problem. Automated help desks are useful tools, but they are limited. "As anybody who has tried to get help from a self-help knowledge base knows, there are times when you can get very frustrated," says Ray Strong, who, with Jeff Kreulen, leads the help desk effort at Almaden. "We want our customers to be satisfied, so we want always to give them a way to get in touch with a human." In fact, HelpNow gives the customer several ways to do so: via IP telephony, text chat or email. The real challenge is getting the customer in touch with the right human. With e-classifier technology, a customer's help request is automatically routed to the e
xpert who has the most suitable skills.
Before HelpNow is rolled out and made available to Global Services outsourcing customers, it is being tried by another group of equally demanding users: the Almaden Research Center population. When Almaden researchers have a computing problem or question, HelpNow will provide them with possible solutions. They'll have the option of initiating an immediate online chat with the adviser whose expertise most closely matches the subject. "The feedback we receive on the system's usability and on the accuracy of the information it provides will aid in improving HelpNow and other new Global Services technologies that we'll put into production here," says Diane Reese, who is leading the introduction of new customer assistance technology at Almaden. The coming year may bring such extensions as new methods of conducting customer satisfaction surveys and better ways to provide solutions from many different knowledge sources. Once all the tools are in use, early in 2000, says Reese, "Almaden will provide a showcase for
what we call the 'Advanced Intelligent Help Desk,' where we can demonstrate technologies in production for Global Services customers."
The central importance of help desks to services has led to various approaches to building Web-based help systems. Service Delivery's HelpNow was developed to integrate the myriad help functions needed to support an entire enterprise's desktop systems. Server users have a different set of problems, and so Product Support Services started working on a project that resulted in IBM Electronic Services, an automated system to support the customer's own server hardware and software. Initially, IBM Electronic Services will support the AS/400®, IBM's popular mid-range computer, but other products will be added in the future.
The automated system addressed a basic need. "There was an unnecessarily large amount of manual labor," explains Sophia Lumelsky, program director of networked business systems at the Thomas J. Watson Research Center. "That caused our CSRs [customer support representatives] to be overloaded and added to the high cost of support and the length of problem resolution time, increasing customer dissatisfaction."
Lumelsky, who led the team that worked on the problem, was able to enlist "knowledge management" technologies similar to e-classifier that had recently been developed at IBM Research. Some of them had not yet appeared in products. The availability of innovative technologies is among the key advantages that accrue to Global Services from its relationship with Research. "Other IT services companies, if asked, 'Do you do research?' would say yes," says Carol Kovac, vice president of services, applications and solutions at IBM Research. "But they would mean market research," and not things like knowledge management.
When a customer logs on to the new IBM Electronic Services for AS/400 Web site, the system automatically pulls up
a description of the latest configuration of the customer's machine. This ensures that the customer will see only
relevant documents. The customer can cruise through the troubleshooting database with a natural-language search that
is far more
powerful than the keyword searches previously used. Every night, the documents in the database -- previous problems
and their solutions -- are sifted by text analysis software from IBM Research, in a
computer equivalent of dreaming. Automatically, new documents are added and new indexes are created.
Often, customers can find a solution by performing a natural-language search on the database of known solutions. If that doesn't work, they can submit the problem electronically and either ask for an electronic notification via the Electronic Services desktop or schedule a callback from a human at a specific time. Once the problem is categorized, the program finds the group of CSRs who have the skills needed to solve that problem. Arranging for the callback to come at the scheduled time is a "very tricky" problem in resource management, Lumelsky says, but well worth the trouble, since it reduces one of the biggest drains on customers' and CSRs' time: currently, 85 percent of the time the CSR calls back, the customer is not there. The scheduled-callback component, implemented by IBM's Haifa Research Laboratory, "is one of the most valuable components of the system," Lumelsky says.
Another time- and money-saver from Research is a collaboration function. This will enable help desk staffers to provide assistance remotely, sharing a Web page with the customer and explaining Web-based reports on performance management, predictive analysis, and alerts.
Customer support is usually a reaction to a customer's immediate, urgent needs. The automated help desk can go beyond that by anticipating problems before they arise. Thanks to a technique known as data mining, says Lumelsky, "we can predict where you're going to be with some resource based on your past and on a segment of customers that share your characteristics." When the customer logs on to the support site, a customized page provides a notification service, warning of any impending problems. "If our predictive analysis says you are going to be out of memory in a month, and we let you know in time to take steps, it would be good business for us, too," Lumelsky says.
The initial pilot tests of the automated help desk were a huge success, according to Ellen Dulberger. "Customer satisfaction went up tremendously," she says. "In fact, when the pilots were over, the customers didn't want to give up the service."
How e-businesses run
During the early months of 1999, IBM Global Services strategist Ellen Dulberger headed a task force charged with analyzing what new challenges and opportunities were going to arise for IT infrastructure services.
The group, whose members were drawn from Research, Global Services, Tivoli (IBM's system management software subsidiary) and Corporate Strategy, soon recognized that the connected organizations that constitute "virtual enterprises" would need new kinds of services, support and management. For example, as IT environments span many companies, systems will need to be administered across the firewalls that were constructed to prevent access from the outside. "This is certainly something Research can help us with," says Dulberger, "and yet, before we began to talk, we didn't even know we needed it."
The task force also began to think about how collecting information on the functioning of these new IT environments could allow Global Services to respond to customer problems faster, or even to solve problems before they occur. The online sale of a book, for example, involves the interaction of many different computers running different software. Outside vendors, such as banks, book suppliers or shippers, might be involved, as well. The customer's credit must be checked, the order must be assembled with other orders from the same region, the shipper must be contacted, and more. Exactly how these tasks are linked and divided -- for example, are the orders consolidated by the bookseller or the shipper? -- can have a tremendous impact on the efficiency of an e-business.
Dulberger and her colleagues saw that by using knowledge management and data mining technologies developed by IBM Research, they could learn how computers actually behave in the real world, learn to anticipate their problems and to diagnose their weaknesses. Instead of focusing their analysis on a single customer or a single enterprise, the researchers could gather and analyze information from new types of environments, from those that encompass multiple enterprises, and from
across environments. That is, they could look closely at the way the information services of many different e-businesses function and learn the rules that govern their health. "Instead of thinking about providing services that optimize one enterprise at a time in isolation," says Dulberger, "we would compare our experience and knowledge base across many enterprises and use that to provide new services in which we predict problems and address them preemptively, to keep e-businesses in business."
A team led by Hamed Ellozy, manager of service management technologies at the Thomas J. Watson Research Center, and a member of Dulberger's task force, is developing the framework for the knowledge management tools that will be used in this new model for multi-enterprise services. The framework will be driven by data obtained from software probes, or agents, that will live in a customer's IT infrastructure. The probes will monitor such behaviors as the load on the central processor, disk drive usage, the number of packets sent and received across the network, the state of applications, business transactions and business processes. Many such probes are commercially available from Tivoli and other vendors.
By gathering and analyzing information from a wide range of e-businesses, IBM Global Services can learn how to help individual businesses improve their operations. While Ellozy expects some companies to be wary of revealing information about their business processes, he is confident that sharing data about how their systems operate will benefit all companies. Having identified the cause of a problem on one customer's infrastructure, Global Services could prevent the same problem from plaguing other customers. Companies could be alerted to the need for, say, more bandwidth or servers, or a change in their network architecture, before such problems began to cripple their business.
Colin G. Harrison, director of global services research at the Thomas J. Watson Research Center, sums up such new services this way: "IBM Global Services is evolving away from just helping to implement and manage collections of boxes and software and moving toward helping to implement and manage e-businesses."
Other types of services will help customers improve the efficiency of their business processes, learning from the behavior patterns of a few and applying them to many. Customers will also be able to see how their systems stack up against the always-changing state of the art. "What e-businesses want most is for their systems to be available and capable of handling workloads that are faster-growing and more volatile than anticipated, and to stay competitive," Dulberger explains. "So if there is a way to keep systems optimized over time, we believe that will have value."
Theory meets practice
Electronic businesses are not just networks of computers, but networks of people communicating by way of computers, making decisions based on the information speeding across their screens. Understanding e-business, therefore, is at least as much a problem of psychology and economics as it is of computer science. A new team of experimental economists at the Thomas J. Watson Research Laboratory are developing tools and theories, and conducting experiments, to help Global Services and its customers understand the new electronic markets and formulate and analyze business strategies.
Insights gained from economic experiments help people like Dinkin and his colleagues to formulate better economic theories, but the participants in such experiments might gain even more from the experience. "They are a discovery tool for senior executives," says Bill Tulskie, a member of the same group as Dinkin. "They are a way for executives to visualize the effect of their business strategies."
Until recently, experimental economics has been pursued mostly in academia and has focused on understanding fundamental economic concepts. Tulskie and his colleagues are scaling these methods up to capture the intricacies of real-world business. "We're trying to tackle problems that are facing management teams today," Tulskie says. "Our sights are set on solving industrial-grade problems while advancing economic theory." To achieve these goals, his group is building a set of software tools that will allow them to rapidly construct, run and analyze experiments.
Another area that Tulskie's group is concentrating on is the creation of tools to represent and assess business strategy. They have helped IBM Global Services Consulting Group develop a tool called a Strategic Capability Network. This, says Tulskie, lets a consultant "sit down with senior executives and help them understand how the resources available to them -- information technology, people, brand -- may be combined to develop organizational capabilities and confer a strategic position in the industry." Using Strategic Capability Networks, a consultant can map the quantitative relationships of a company's value propositions -- the kinds of things they might put in an ad, like "lower prices" and "faster service" -- to its ability to deliver on those propositions and
to its underlying assets. Capturing a snapshot of a company in this fashion helps executives analyze the effects of changing different aspects of the business model. For example, a Strategic Capability Network can unravel the difficult problem of determining how well the business models of two companies might be able to merge.
The tool is particularly useful for established, brick-and-mortar corporations looking for an edge against dot.com competitors, says Michael Shank, a principal in IBM Global Services' e-business strategy and design practice. Shank uses the example of a traditional book retailer under siege by online competition. The online vendor might offer personalized book recommendations based on its knowledge of online buying patterns. But the traditional bookseller might have an edge in its face-to-face customer interactions -- interactions that could yield even richer knowledge of customers, if the bookseller could only capture it. "The Strategic Capability Network shines a light on these kinds of hidden capabilities, revealing their potential business value," Shank says.
IBM's e-business strategy team has put the tool to use for customers in a range of e-business initiatives. In many cases, managers in one business unit were unaware of initiatives underway in another. In other cases, projects were scrapped because investment costs were too high. The Strategic Capability Network takes an enterprise-wide view, enabling business executives to understand how various initiatives can be integrated to give businesses more for their money. "The Strategic Capability Network has had a huge impact on our clientele," says Shank. "We were able to commercialize this work and take it to customers almost immediately. We are now accelerating our deployment of this tool in our consulting engagements around the world."
Tulskie and his group are using the same concepts to study past e-business initiatives, particularly ones aimed at extending a business. "I think that is the key to what firms want IBM to bring to them," Tulskie says. "An expression of what other firms have learned from experience. What sorts of things did they build in their Web catalog initiative? What sort of intranet systems did they build? What policies and practices did they accumulate on the way? And firms want answers that are usable in their own businesses." IBM Global Services Consulting Group and Watson researchers are conducting a benchmark study to explore how the Internet and related technologies are helping businesses. Such studies, combined with the "industrial-strength" tools developed by Tulskie and his colleagues, are helping to bring the insights of modern economic research from academia into the real world.
Knowledge portals
The importance of knowledge in the services business is hard to overstate. "The differentiators that you deliver to your customers are knowledge and expertise where and when they're needed," says Carol Kovac, the IBM Research vice president of services, applications and solutions.
When Global Services was small, locating sources of knowledge and expertise was relatively simple. "But now, making knowledge available to the right person at the right time in the right language is a problem," says Robert Mack, a member of the text analysis and advanced search group at the Thomas J. Watson Research Center. "That's because the people at Global Services deal with all kinds of customers, and a lot of what they do builds on what they've done in the past." The information generated by an engagement with a customer -- the results of meetings, sales calls and strategy sessions -- is spread out over many repositories: file servers, Web sites, email, Lotus® Notes® databases and elsewhere. "This constitutes potential knowledge," Mack says, "knowledge that can lead to more effective customer engagements."
To bring that potential knowledge to life, Mack used advanced knowledge management technology developed at IBM Research to create a knowledge portal for e-business information.This technology includes text analysis and search, as well as document categorization technology developed at Watson. An expert in a particular area defines a set of categories describing that domain of knowledge and then identifies a set of training documents that exemplify each category. By training on this set, the system determines the key features of documents belonging to each category, and classifies subsequent documents accordingly.
The classifications represent the way consultants think of their data. For example, one category is "engagement life cycle" -- all the stages in working with a customer, from proposal to solution. The documents that are categorized are identified, in turn, by a special "crawler" technology called Grand Central Station, developed at IBM's Almaden Research Center. The crawler collects documents from different sources, including the Web and Lotus Notes databases, and in different formats, such as spreadsheets, slide presentations or reports. The categorization software sorts all this disparate information.
To make the Research tools, as well as the tools already used by Global Services, accessible even by modem, Ko-Yang Wang, manager of knowledge management application development for IBM Global Services, developed a Web application that reduces the information that must be downloaded. A recent field test by 900 IBM Global Services consultants showed that the portal can indeed find useful documents.
The partnership between IBM Research and IBM Global Services is predicated on businesses' insatiable need for knowledge. "IBM Research is a knowledge factory," declares Colin Harrison, Watson's director of global services research. That knowledge, whether gained in the laboratory, through data mining or through economic experimentation, is helping Global Services to provide business solutions more efficiently. "What Global Services traditionally offers customers," says Harrison, "is a pair of skilled hands plus the intellectual property that comes from training and work experiences. In the past, most of the value has been in the hands. But through their collaboration, Global Services and Research are shifting that balance toward knowledge."
Bruce Schechter, a freelance writer who lives in Brooklyn, New York, is the author of My Brain Is Open:
The Mathematical Journeys of Paul Erdös.
