– from venture startup to core contact center product by Glen Taylor
I’m not the best person to tell this story. I came in during a period of transition and was not a key player in the evolution, although I was affected by the winds of change. However, as I began to build this website, I became interested in understanding the story: how did the Conversant venture come to be and how did it evolve into a successful product in the AT&T, Lucent, and finally Avaya contact center portfolio. The following is what I learned.
In 1982, AT&T had settled an antitrust lawsuit with the US Department of Justice. AT&T had agreed to sweeping organizational change – the “breakup of the Bell System” – in exchange for some relief that it felt would propel the company forward. In the 1950s, AT&T had agreed in another antitrust decision to refrain from competing in the computer business. The result of that agreement was that some incredibly significant computer science innovations, for example the Unix operating system and the C programming language, could not be directly sold by AT&T. Yes, they were the basis of much of the innovation and development done by AT&T Bell Labs, but they were “hidden” within AT&T’s products and services – the secret sauce. Since AT&T could not “sell” computers, UNIX and its tools, or the C programming language, they were made available to the academic community for the price of a computer tape and the labor to copy it. Ironically, by seeding the academic institutions that trained the next generation of computer scientists with this powerful software, it would change the computer business in fundamental ways making Unix and its derivative Linux the dominant operating system. This “freely available” OS would ultimately displace the commercial offerings of IBM, Honeywell, Burroughs, Data General, Digital Equipment Corporation, and other major computer vendors most of whom no longer exist. It also initiated the concept of openly sharing significant software source code and is therefore arguably the genesis of the “open source” movement so important today. However, I digress.
One of the benefits that AT&T got from the 1982 agreement was the ability to compete directly in the computer business. It also was no longer just a regulated monopoly. Oh, it retained its monopoly business, but it could own and operate unregulated subsidiaries. Hence, the creation of American Bell in July 1982. The winds of change were beyond gale force on the monopoly telecom business, but AT&T was trying to meet that challenge by becoming more of a competitive, unregulated entity. In the spirit of that time, someone within AT&T hit upon the idea of creating an internal venture funding unit. Some of the boundless creativity of the Bell Labs community could be tapped to develop novel products. Individuals with good ideas could apply for this internal funding and create an “intrapreneurship” – an “internal venture.” A small business, effectively a startup, was carved out of an existing organization. The employees remained in their current titled positions within (primarily) Bell Labs or another AT&T entity, but they had a different role getting the venture’s product created and ready for market. If their effort was successful, AT&T would either find a place in its portfolio to sell the product or it would spin off the venture and retain an equity stake in the business.
When I arrived at the AT&T Consumer Products Bell Labs unit in Indianapolis in 1983, there was an existing internal venture that would be successfully spun off in another one to two years. In late 1985 or early 1986, I became one of about fifteen Bell Labs personnel that began another internal venture. It wasn’t related to IVR, and it didn’t make it out the door. However, the reason it failed to launch wasn’t technical. By 1986, AT&T was undergoing serious reorganization at headquarters. The winds had shifted direction. The people involved with venture funding realized that their time was over, and they left AT&T. Venture funding dried up and all of the ventures, including the one I was involved with, were left to wither and die or fend for themselves. That was, in part, the reason I joined a cadre of Bell Labs MTS who transferred from Indy to Columbus the first working day of January 1987.
As mentioned elsewhere, Dean Hester supervised a group in Tony Cuilwik’s department. Dean’s group was responsible for developing the #2 Service Evaluation System, an operations support system administratively within AT&T Network Systems and designed to do quality sampling of long-distance calls. Dean’s group contained several individuals that I would recognize as “key players” and “originals” when I joined the Conversant organization in January 1987, but the original work took place around 1983 or 1984. One of Dean’s top Member of Technical Staff (MTS) was Bob Perdue. Bob had done an internship in Larry Rabiner’s research organization in Murray Hill, NJ, on signal processing and speech recognition. The #2 SES was already using some fairly sophisticated signal processing to recognize various telephone call conditions based on the sampled auditory information. Adding speech recognition capability was a next step, but was something more possible? One Saturday morning, Dean called Bob with an idea. Could they build a voice response system that would allow people to interact with it over the telephone?
Dean was apparently successful convincing Tony, his department head, that he had a good idea. Dean’s group would embark on something affectionately called a “skunk works” within Bell Labs. Although they were still supported by their parent organization, they were building something unrelated to their parent organization’s charter but with management approval. This activity would be called Speech Processing Products. There was another bright, energetic Bell Labs department head with an entrepreneurial spirit, Kendra VanderMeulen. How Dean and Kendra began to work together, I have not been able to determine, but they were a perfect pair to move the venture forward. They effectively became a separate Labs organization with Kendra promoted to be Director (President of the venture) and Dean as R&D Department Head. Two additional key individuals, Sandra Oliver-Sterbenz and Harry McHugh, would be enlisted to be the head of operations and the head of sales & marketing, respectively. Kendra would name the unit AT&T Conversant Systems. The effort moved forward successfully creating the initial product, the Conversant 1 Model 80. There was significant effort to find some early customers, get trials done, and begin to roll out the product. That first product was declared “generally available” in November 1985.
Conversant was beginning to show promise as a very functional programmable voice response system, what we would come to call interactive voice response or IVR. There were other IVR systems commercially available, but this was AT&T’s own, organically created product using internal AT&T manufactured components and intellectual property. However, the path to success was not straightforward. Conversant Systems was carved out of a group reporting up to AT&T Network Systems, the portion of Bell Labs that supported the AT&T long distance telephone business, the part that would remain regulated. It was manufacturing a product in the Columbus Works Western Electric facility, but Western Electric didn’t sell to the general public. There was a market for this technology within the AT&T network, but to be commercially successful as an IVR product, Conversant would need to become widely sold to other enterprises. The mis-alignment between these factors was problematic while Conversant was being venture funded, but it would become much more problematic when that funding source disappeared. By the end of 1986, the Conversant Systems’ management and staff were looking for customers wherever they could find them. It was probably unclear how much longer the Network Systems organization and Western Electric would be willing to “lend” personnel to a project that wasn’t a clear fit.
Anyone who worked in Bell Labs during the 1970s and 1980s will probably acknowledge the “not invented here” attitude. A term that usually meant “you may have a good idea, but I can do it better.” Bell Labs was large and the Bell System larger. Every component of AT&T and the operating telephone companies, who were now “customers” of AT&T, had needs that could be traced back to some organization within Bell Labs whose job was to support them with products, procedures, etc. This distributed structure led to considerable duplication of effort. Even if two organizations recognized that they were each tasked with doing effectively the same thing, they couldn’t easily pool resources and consolidate. They answered to different corporate management with different goals and timelines. Add to this the fact that the very bright people who populated Bell Labs and its management had comparably strong personalities and large egos. All of this led to competition and various forms of in fighting. Such was the case with Conversant.
One of the first successful applications of the Conversant product was in the guise of an Outbound Call Management (OCM) system, an out dialing platform used for debt collection. American Express was a key target customer as were some AT&T business units. OCM was technically “owned” by an organization in NJ, but was effectively the Conversant Model 80 scripted to do out dialing. That would create a political tension between NJ and Columbus. As a result, the NJ organization decided it could build its own platform and do it better. A group in Middletown, NJ, was creating a voice card they called the Voice Power board that would fit in a PC. The group who “owned” OCM was planning to use that Voice Power card to create a platform they called the Voice Information System. There was significant political tension between Columbus and NJ over this. Kendra and Dean were politically skillful and within a year, the Voice Power card had been transferred to Columbus and Conversant had “absorbed” the Voice Information System effort changing its name to Conversant VIS.
Shortly after I arrived in Columbus, my supervisor, Sarbmeet Kanwal, and I would travel to Denver to meet with the voice messaging Audix organization and to NJ to meet with Bell Labs systems engineers supporting AT&T Long Lines Consumer Communications Services (CCS). At the same time, others in our organization were working diligently to secure a deal with an American Express – AT&T joint venture located in Omaha, NE, First Data Resources – Interactive Technology. Yes, these were “sales” activities heavily involving the Bell Labs team. Conversant needed sales and it needed funding support.
I would end up being interned to that NJ systems engineering group I’d first visited in early 1987. I would successfully “sell” them and their CCS customer a Conversant-based architecture for the store & forward messaging service CCS wanted to offer its long-distance customers. That was the Call Delivery Service project that would become the trademarked AT&T VoiceMark Messaging offer. This initial deployment of Conversant with CCS would lead to opportunities with Long Lines Business Communications Services (BCS). BCS had an offering for its largest customers called the Software Defined Network (SDN). SDN used the public telephone network to effectively provide large enterprises was a private telephone network. A feature of SDN was Network Remote Access (NRA) whereby the enterprise could issue a special calling card to its employees. With the card’s authorization code, long distance calls could be charged to the enterprise’s SDN account. But, BCS had a problem with SDN NRA. Users could not conveniently make multiple calls in a single interaction which was very inconvenient. I proposed an architecture that placed Conversant in the middle of these calls enabling it to give callers sequence calling. This SDN NRA Sequence Dialing project positioned Conversant in the middle of the AT&T long distance network. We demonstrated that using Conversant we could rapidly develop a network feature that would have required years of development within the Network Systems core network elements.
Later, I would also be deeply involved in the application of Conversant technology to AT&T operator services, a project known as Voice Recognition Call Processing that was reputed to have saved AT&T hundreds of millions of dollars annually. All of these projects are documented elsewhere, but they brought in not only significant CONVERSANT sales but “work for others” funding from the parent AT&T organizations that paid for our efforts in Columbus.
Although Kendra and Dean had success navigating the political waters, creating a product that would be an industry leading IVR, their hold on the product was slipping probably as early as 1988 or 1989. It was becoming increasingly clear that the future success of Conversant could not depend on internal funding and internal AT&T applications alone. It needed to be sold to enterprises and it needed to fit into their contact center infrastructure. That meant that the product needed to move from Network Systems to Business Communications Systems (not to be confused with Long Lines BCS division).
Starting in the mid-1980s and following a timeline similar to the one outlined here, Business Communications Systems (soon to be called Global Business Communications Systems or GBCS) had grown up. As a regulated entity, AT&T owned the telephone network and the switching systems that directed network traffic. Large enterprises, however, had enough internal telephone stations that they justified purchasing and operating their own small switches, the private branch exchanges or PBXes. As call volumes into businesses grew, so did a structure for dealing with those calls, the call center. A PBX would include automatic call distributor or ACD software that directed calls to different functional groups answered by separate pools of agents. In order to give some sophistication to the distribution of calls, the PBX ACDs had developed the ability to play simple messages, collect a touch tone digit, and based on the caller’s input, direct the call appropriately. Although these ACDs could do this simple call routing, call flow configuration was awkward and limited. To provide more sophisticated interactions with callers, an IVR platform was required.
By the mid-1980s, AT&T’s two flagship PBXes, the larger System 85 and the smaller System 75 were “competing” with each other. System 75 development was primarily in NJ with System 85 development primarily in Denver. Each platform and each development group had unique strengths and features. The resolution of that competition would be a joining of forces and the declaration of a new combined switching platform, the Definity series. Initially, that was just a renaming. Definity G1 was just a System 75 renamed, and Definity G2 was just a System 85. It would take some time before the actually merged Definity G3 emerged. GBCS was the AT&T corporate entity that provided product management direction to this line of products and was incorporating related technologies used in the call center. Those would include reporting with the Call Management System or CMS, voice messaging with the Audix system, and eventually IVR with the Conversant system. However, our story hasn’t gotten there yet.
By 1989, the political pushing and shoving to realign Conversant, wresting it from Network Systems and aligning it with GBCS, was taking place at higher corporate levels. The concept of Conversant as an “internal venture” was effectively at an end. It would be a supported call center product, but not as a small separate business unit lodged somewhere in the larger corporate structure. The business aspects of the Conversant venture – product management, marketing, sales, manufacturing, and support – would all be realigned. The GBCS product management organization, primarily located in NJ, would become Multimedia Messaging and Response – a combination of voice messaging with interactive voice response. Marketing would move to the corporate level. Sales would become the province of the AT&T GBCS sales teams who sold the Definity switches, the Contact Center Elite ACD software, the CMS reporting product, the Audix voice messaging product, and now the Conversant IVR product. Manufacturing would remain partly in Columbus, but much would be moved to the Denver Works. Support for the product would migrate to the GBCS support organization located in the Denver area. The realignment within Bell Labs would be to combine the Audix development department in Denver with the Conversant development department in Columbus under Dean Hester’s then Bell Labs director, Charlie Del Riesgo. There would be more evolution after this point but the movement from a Bell Labs “skunk works,” to successful “internal venture,” to GBCS call center product was effectively over. Before Charlie became the Conversant Bell Labs director, Kendra had departed becoming the director of a different Network Systems Lab in Columbus. Dean would soon move to take a different department in Network Systems. Conversant would be a GBCS island within the Columbus Network Systems Bell Labs location. In Feb. 1991, Charlie Del Riesgo retired from Bell Labs and moved to Cincinnati Bell Information Systems in Cincinnati. Kendra and a number of other managers and MTS would ultimately follow, but that’s another story.
