 |
 |
 |
 |
| |
|||||
|
|
|
Home > News & Events > MITRE Publications > The MITRE Digest > | |||||||||||||||||||
Testing the Future of IRS Customer Contact March 2003
Handling more than 100 million customer phone calls in six months could cripple most businesses, even L.L. Bean or Dell Computer. Yet the Internal Revenue Service (IRS) does that during high tax season every year. And its "catalogue" isn't a finite list of product information, but a set of ever-changing tax regulations filling thousands of pages. Faced with a call volume and level of complexity unheard of in the commercial sector, the IRS is working hard to improve its customer-service technology, especially at its front-line call-centers. As one of the IRS's partners in its Business System Modernization, MITRE's Center for Enterprise Modernization (CEM) has created a specialized laboratory to bolster research of contact center infrastructures. This research is key to discovering worthwhile ways government agencies can implement automation to improve customer service and reduce operations costs. In what they call the Contact Center of the Future, or CCOF, CEM's researchers investigate state-of-the-art telephony, e-mail, and Web-based technologies that have the potential to improve taxpayer-agency customer communications. The knowledge gathered by this research effort also applies to similar initiatives at other agencies—such as the Department of Veteran's Affairs (which CEM also supports) and the Social Security Administration—that have to communicate effectively and conduct business with internal or external customers. Meeting High Expectations "Customers expect 24/7 customer service, and they want quick, accurate information without going through 'touch-tone hell,'" says David Madison, a CEM principal information systems engineer. "Yet running call-centers is very labor-intensive, and many of them need modernization. That's not unique to the IRS, but the IRS and other government agencies do have unique challenges that differentiate them from commercial companies. For instance, the IRS's call-center work is very peak-demand oriented, specifically between January 1 and April 15 of each year. You need to build infrastructure to support that peak, but without spending so much that you have enormous excess capacity the rest of the year. Also, there are special privacy issues. Tax and social security data are very sensitive and must be secured." Managing a vast amount of information is another hurdle the IRS faces, given the sheer size of our nation's tax code. Each of the approximately 10,000 customer-service representatives has access to different information; no one can be a specialist in every detail of the U.S. tax code, so calls have to be directed to the right expert. "We also must consider the different channels for getting and receiving information, including phone—both speech and touch-tone menus—human interaction, mail, and the Internet. Providing consistency of customer service, regardless of the channel, is a clear challenge," Madison says. "We believe MITRE can help." Testing the Future As the manager of the IRS's Federally Funded Research and Development Center (CEM), MITRE is integrally involved in the agency's enterprise-wide, multiyear Business Systems Modernization program. One of the program's main goals is to provide "top-quality service to each taxpayer in every transaction." Through a combination of automated and human-assisted responses, service has steadily improved since the modernization effort was initiated a few years ago. But the IRS still wasn't satisfied with telephone response rates and wait times. "MITRE got involved in this specific area through the IRS's Customer Communications 2001 Project, which is part of the IRS's modernization program," explains Paul Herceg, a lead information systems engineer with CEM. "The project focused both on enhancing ways to accept and distribute phone calls and on adding a new voice-initiated refund-status software application, in English and Spanish. The goal was to update the IRS's telephone systems and bring additional capabilities to bear." The IRS activated this enhanced phone system in the summer of 2001. Even with these improvements in the works, the MITRE team envisioned ways to make the system more efficient and cost-effective, with ideas such as integrating non-voice communications (including e-mail and Web applications) and state-of-the-art speech technology into future contact centers. Herceg contacted MITRE colleague Susmit Patel and proposed a cross-collaboration between CEM and MITRE's Department of Defense FFRDC (where Patel is a principal engineer). This partnership helped make the Contact Center of the Future project a reality. Using a combination of advanced telephone system hardware and software, the lab acts as a testbed for experiments and demonstrations. First up: determining how to best route each call as it comes in and send it to the right customer-service representative. Through collaboration with Margot Peet, a principal engineer in artificial intelligence, and her colleagues in MITRE's Signal Processing Center, CCOF researchers have been working with Nuance, a company that specializes in voice recognition and computerized telephone applications, to aid in the experiments. "We want to automate the process as much as possible, without turning off the customer," Madison says. "Figuring out how to route calls is a major challenge." Leveraging Open, Standards-based Infrastructures MITRE experts believe that converging call centers onto an open, standards-based infrastructure may offer improvement opportunities for agencies such as the IRS. To demonstrate this, we built the CCOF lab using open, standards-based technology, such as Voice-over IP (VoIP) and VoiceXML—short for Voice Extensible Markup Language. VOIP enables a phone call to be handled on a standard all-purpose data network. VoiceXML is a standard language for creating voice-user interfaces, which can support phone calls. It uses approaches and structures similar to those for Web development. "Most call centers use closed, proprietary telephone systems," Madison says. "By converting incoming phone calls to run on all-purpose data networks, all aspects of a call-center can be integrated -voice, e-mail messages, Web contacts, and so on. You want the incoming calls and Internet questions to 'look' as similar as possible so that you need just one infrastructure to handle everything and to provide a consistent response. Conceptually, the idea is to merge all channels into a single communications network." Although companies such as Cisco, Avaya, and Nortel currently offer VoIP- and VoiceXML-based products to their customers, few businesses use an open, standards-based solution for their entire customer-contact center operation. Because technology based on such standards is maturing, part of MITRE's CCOF research is designed to help determine the pros and cons of applying each technology to the call-center environment. An advanced call center's systems would funnel all incoming message traffic—whether voice or text—into a common digital network, which would then sort out the caller's intent and leverage automation to answer inquiries or conduct transactions. Intent identification is one of the key factors in properly routing a question to the correct customer-service representative or automated system. To the extent possible, an efficient system would recognize the customer's intent without a human in the loop, although many calls/questions would be routed to a person. Questions needing a straightforward, data-driven answer such as "What's the status of my account?" would go directly to an automated system for a response. Having everything in an open, standard format will also make it easier to capture inquiry data (such as the types of most commonly asked questions) and analytic data, used by managers and executives. This information would be used to improve future customer-service initiatives. Advanced Customer Service for All One of the advantages of the CCOF lab is its ability to demonstrate cutting-edge technology, both other companies' and our own. For instance, in addition to the Nuance speech recognition software, the researchers have been testing a "conceptual browsing tool" for searching IRS publications and running experiments using Qanda, a MITRE-designed question-answering system originally developed for the Department of Defense. As Herceg and Madison continue their research, they are exploring ways to take advantage of software that recognizes words related specifically to the call center's function. Language processing is a key to future intent identification and automated customer service systems. Machine translation and language understanding are areas of significant interest to the CCOF project. It's not as simple as it sounds, however, the researchers say. The problem isn't so much recognizing words like "return" or "exemption," but understanding the semantics of a word in a given context. For instance, the utterance "Mr. Wright should write to Ms. Wright right away about his tax return" poses a considerable challenge to a voice recognition system. While these improvements are sure to improve customer service, Madison and Herceg say the cost benefits of converting contact centers to an open, standards-based infrastructure have yet to be proven. Thus much of this technology needs to mature before agencies make significant investments. Government agencies such as IRS operate under severe budget constraints, making the choice of how to spend their technology dollars a critical one. "We want to be in a position to help our customers look into the future, lay out their technology roadmap, and avoid technical dead-ends," Madison says. —by Alison Stern-Dunyak Related Information Articles and News
Websites |
||||||||||||||||||||
| Page last updated: April 12, 2004 | Top of page |
Solutions That Make a Difference.® |
|
|
