The Voice of the Future

by Michael Peltz

February 2001, “Worth” magazine

At a time when no one wants to hear about the next big thing, Mike McCue may be building just that. It’s called Tellme — and to use it, all you need is your phone.

Have you noticed that those who are in a race to be first are often late? Tellme Networks CEO Mike McCue is certainly no exception. McCue, who was vice president of technology at Netscape Communications, has been sprinting since he cofounded Tellme in February 1999; he is so passionately engaged in what he’s planning or whom he’s talking to that he usually has a hard time calling it quits. I knew this about the 33-year-old McCue as I waited for him outside New York City’s ’21’ Club following a press luncheon this past November. The purpose of the lunch was a briefing about 1-800-555-TELL, the toll-free voice-activated service that McCue’s company had recently launched.

Don’t be misled by the prefix. Even though Tellme might seem, on the surface, a garden-variety 800 number, it is anything but. In fact, it may just be the transformative leap that many have argued the Internet is waiting for, a technology that wrests access away from the computer and delivers it through the humble ubiquity of the telephone. That’s right: no computer, no DSL line, no modem, no BlackBerry, no snazzy Internet-enabled status symbol needed. Tellme allows anyone to obtain information, businesses, and services over the Internet by just speaking into a telephone receiver.

I had been following Tellme closely since last spring when AT&T (NYSE: T) invested $60 million for a small stake in the Mountain View, Calif., company. I became even more interested when Tellme raised, with relative ease, an additional $125 million this past October, months after virtually the entire field of Internet startups had gone from heroes to zeros. With this market-defying performance in mind, I was looking forward to tagging along with McCue and colleague Hadi Partovi as they drove to Basking Ridge, N.J., for a meeting with the top brass at AT&T’s consumer group. The trip might help me understand how these two entrepreneurs, once dire enemies, had found common ground in their almost religious zeal to make the Internet work for people instead of the other way around.

“We must have met with 50 journalists in the past two days,” McCue gushed when he finally emerged from the restaurant and we piled into the backseat of a black Lincoln. “And everybody really got it.”

I wasn’t surprised. While so many new-economy businesses strain the brainpower of anyone to understand them, Tellme’s story can be instantly understood. Its pitch ó based on delivering data and information such as stock quotes, sports scores, horoscopes, movie times and driving directions over the phone, using speech-recognition technology and the Internet, is richly intuitive and hence deeply impressive. Looking for a restaurant in a strange city or an unfamiliar part of town? Not only does Tellme help you zero in by neighborhood and cuisine, it’s thoughtful enough to provide a Zagat review, if available. It can then connect you to your choice so you can make a reservation. Need a taxi after dinner, or want to catch a movie? Tellme obliges.

In bridging the Internet and the telephone, Tellme fills a huge gap. Depending on how you look at it, the service could be variously described as an Internet voice browser for the phone, a phone browser, a voice portal, or a phone site (as opposed to a Web site). Yet behind the seductive consumer front end are icebergs of technical challenges, including the difficulties inherent in speech-recognition and text-to-speech technology, which Tellme licenses rather than owns, as well as constructing a network of servers and phone lines to link the parallel universes of the Internet and telephony. When you call Tellme for the name of a restaurant in New Orleans or for a taxi in New York, this simple request involves an Internet search for the information and uses technology that instantly converts that text into speech and delivers it in a relaxed, friendly voice, instead of the robotic, synthesized speech heard on a typical voice mail message. Beyond these technical hurdles, Tellme needs to find a way to capitalize on its investment by selling advertising on its phone site, charging commissions for referrals, or some combination of both.

Recognizing the challenges in achieving profitability based on the consumer applications of Tellme, this side of the business, grand as it might be, is just part of McCue’s hugely ambitious plans. The second component is a business-to-business application: Tellme’s voice-enabled network platform ó which powers its directory and sits between the traditional circuit-switched phone network and the Internet ó can also replace all the infuriating, user-unfriendly embedded voice-mail applications that thousands of businesses are saddled with. These businesses can turn to Tellme for the outsourcing of all the infrastructure required to answer any phone call with a superior, friendly, and economical voice-activated application (ultimately, in any language). Think travel reservations, mail order, customer service, stock trading.

The revenue implications of this are staggering. Ken Hyers, an analyst at Cahners In-Stat, a research group, expects the voice-portal market to reach $5.3 billion in annual revenue by 2005 (with $1.6 billion coming from advertising, sponsorships, and e-commerce and $3.65 billion from reselling marketing profile data). Kelsey Group analyst Mark Plakias uses a much broader definition of voice-related services. He forecasts that Tellme’s market could reach $26 billion globally by 2005.

Tellme’s potential was impressive enough to secure its latest capital infusion at a time when consumer-focused Internet businesses were putting venture capitalists into anaphylactic shock. With consumer and B2B companies anxious to cut costs while developing better customer service, we’ve all read about the obsessive focus on CRM (customer-relationship management), the applications that Tellme offers are in demand.

The picture for investors remains somewhat sketchy. Tellme is a private company, though McCue may take it public this year. The technology it is developing, however, does promise to bring about broad change in the way business gets done. And whether Tellme ultimately prevails in this arena, this change will impact a wide array of stocks, for better and for worse. Certainly, e-tailers such as Amazon.com (Nasdaq: AMZN), which reportedly has been working with Tellme to build a voice-activated phone application, stand to benefit by increasing their reach. The technology could also prove to be a boon for one of America’s oldest and largest corporations, AT&T, or at least what’s left of it.

En route to the communication giant’s operational headquarters, McCue pulled out his cell phone to check in with Emil Michael, Tellme’s head of business development. Partovi booted up his laptop and was trying to read e-mail, not an easy task as the speeding Lincoln wove in and out of traffic on Fifth Avenue. It felt like a scene out of last year’s “Almost Famous,” Cameron Crowe’s 1970s rock-band-on-the-road bio-epic, except the whiskey was replaced by wireless modems, the drugs by digital devices.

McCue fervently believes that Tellme’s voice-enabled applications will have a powerful twofold benefit for carriers such as AT&T. Customer-acquisition costs are high and getting higher; Tellme’s technology will improve user satisfaction, reducing customer churn rates and increasing retention. The economic advantages of this are astounding in the aggregate. Tellme’s applications will also boost usage minutes. That should help offset the huge drop in long-distance calling rates resulting from deregulation and increased competition, not to mention the emerging threat of Internet telephony.

Amid this revolution, part of the familiar soundtrack of everyday life will disappear. Or so McCue believes. Ultimately, he told me, the dial tone as we know it will die. “When you pick up a phone,” he explained, “you’ll hear a friendly voice say, ‘What would you like to do?’ and you’ll be able to place a call or do a whole variety of things using simple key words.” This future ó which McCue addresses with evangelical fervor, is what he refers to as “dial tone 2.0,” and he expects it to start to become a reality within the next two years. The impending breakup of AT&T into consumer, wireless, broadband and business services has only accelerated the dial tone’s eventual extinction.

McCue really is trying to change the world, and his enthusiasm is infectious. His tech pulpit skills have enabled him to assemble a world-class team of engineers, and bring together an impressive list of financial backers. In total, privately held Tellme has obtained $238 million from investors since July 1999, and its $125 million financing round this past fall was larger than the capital raised in most of the year’s IPOs.

McCue moved to Silicon Valley in 1996 after he sold his Paper Software, based in Woodstock, N.Y., to Netscape for $20 million. Joining Netscape was part of the deal, and it provided a chance to work with Marc Andreessen, who had developed the original Mosaic Web browser. One of McCue’s first assignments was a project he called Constellation (named after a historical boat his father helped restore when McCue was growing up). The goal: to transform Netscape’s then market-leading Navigator into a Web-based desktop operating system. The story goes that it was the threat of Constellation that led Microsoft (Nasdaq: MSFT) to alter Windows’ licensing agreements with PC makers to prevent their using competing software products. Effectively blocked, Constellation never got off the ground. The episode eventually became one of the triggers for the Department of Justice’s antitrust investigation of Microsoft. (When McCue later bought a $200,000 48-foot classic wooden sailboat, he named it “Constellation.”)

As fate would have it, McCue’s counterpart at Microsoft during the DOJ case was Hadi Partovi, who headed the Internet Explorer team. Partovi first met McCue and his Netscape colleague Angus Davis during the 1997 Comdex trade show. One night, after a day of walking the endless convention floor, their group went out and ended up at a karaoke bar, singing “You’ve Lost That Loving Feeling.”

McCue credits Davis with the original idea for Tellme. Davis was a summer intern at Netscape when he caught McCue’s attention with a talk on how to build a Web site that would work with any browser. Afterward, Davis launched into a techno riff in which he peppered McCue with reasons that Netscape should consider using the telephone as an interface device. “Everyone has access to the phone,” Davis reminded him. “How cool would it be if you could access the types of things you access from Netscape’s servers over the phone instead of just over the Web?”

McCue took to Davis right away: “Why are you going back to college? Why don’t you come work for me instead?” What McCue didn’t realize was that Davis, then only 19, hadn’t even started college at the time. McCue himself had gone directly from high school to IBM (NYSE: IBM).

McCue and Davis didn’t talk much about what would become Tellme until the end of 1998. By then, the great entrepreneurial ride of Netscape had already come to an end: America Online (NYSE: AOL) had agreed to buy the company, and McCue was starting to feel blocked and underappreciated. “I had gotten the reputation for not being an operational guy,” he says, adding that he would have stayed on if his request “to give me Netscape Navigator and let me make it into something really great” was honored. It wasn’t, and in December 1998, McCue left to start Tellme. Within a month, he had convinced Davis to join him.

McCue also sent an e-mail to Netscape cofounder and serial-start-up artist, the billionaire Jim Clark. McCue’s vision intrigued Clark enough to wrangle him an invitation to Clark’s house in Palm Beach, Fla., the very next weekend. The two brainstormed the idea, pushing its limits, stress-testing it, was it dial tone 2.0, interactive radio or some combination of both? The exchange continued through rides in Clark’s helicopter, stunt plane and Ferrari. By the end of the weekend, McCue had a tacit commitment from Clark, a development that, because of Clark’s wide influence, virtually assured McCue would be taken seriously in money circles.

Tellme Networks was incorporated by McCue and Davis on Feb. 16, 1999. In early March, the two partners held a founders’ weekend, bringing together 25 of the best and brightest engineers, programmers, and speech-recognition experts in their collective Rolodexes, PDAs, actually, as stage one of building a team. To illustrate Tellme’s potential, McCue and Davis ran a crude demo they had prepared that showed how, using a phone and simple speech-recognition commands, they could find a nearby restaurant over the Internet and order pizza, the primary food group for Silicon Valley startups. Danny Rimer, then an Internet analyst with Chase H&Q, and Gigi Brisson, a late-stage private-market investor, also appeared at the meeting to share their wisdom about the potential of the voice Web market. (Brisson’s firm, Attractor, led the most recent funding round for Tellme; Rimer, now a partner with the Barksdale Group, was Tellme’s first investor ó he wrote McCue a check for $25,000 when McCue told him he was starting a new company.)

McCue and Davis convinced 15 of the 25 potential recruits to sign on as Tellme’s original founding team; five of the remaining 10 would later join as well. Hadi Partovi was too wrapped up in the launch of Internet Explorer 5.0 to commit right away. But McCue remained persistent, and some of his phone calls and e-mails got through to Partovi on his postlaunch vacation in Spain (Partovi never goes anywhere without his laptop). He joined McCue a few weeks later.

The unlikely combination of erstwhile rivals McCue and Partovi enabled Tellme to attract a powerful blend of talent from Netscape and Microsoft. It was also critical in the company’s initial July 1999 financing, which brought $6 million in seed money from a small group of wealthy individuals led by another pair of ex-enemies: former Netscape CEO Jim Barksdale, who was just setting up the Barksdale Group; and former Microsoft senior vice president Brad Silverberg, who would later form his own investment firm, Ignition. Both put their differences aside to participate in Tellme, says Silverberg, who sits on Tellme’s board.

Where was Jim Clark in all this startup check writing? Nowhere to be found. The Netscape co-founder decided to pass on the deal the very morning he was supposed to wire the funds, because he was too busy with his own latest start-up, myCFO. “I was shocked,” says McCue, who ended up offering Clark’s piece to Barksdale. “The deal could have easily unraveled, but it actually turned out to be one of the best things that happened to the company. ‘We can do it without him’ became a rallying cry for us.”

The science of speech recognition and speech generation isn’t new, nor is the dream, remember HAL from Stanley Kubrick’s 1968 classic, 2001: A Space Odyssey? Ironically, the driving force behind the technology, and thus behind Tellme’s future, sprung out of the clumsy, user-unfriendliness of most computer interfaces. As Nicholas Negroponte predicted in “Being Digital” back in that prehistoric year of 1995, the goal for a computer is “recognizing and understanding conversational speech in a highly personalized and interactive environment — from, say, Fran Drescher to William F. Buckley Jr., requires millions of computations. Today, software programs from companies such as Nuance Communications (Tellme’s speech-recognition supplier) and SpeechWorks have large vocabularies and can learn to recognize a wide variety of speech patterns. Similar advances in text-to-speech software have improved a computer’s ability to “read” just about any text and convert it to voice. “The technology of speech recognition and text-to-speech has improved tremendously,” says Mark Winther, head of the world-wide telecommunications group at the research firm IDC. “It is very good now.”

But the killer app is not just accuracy. Accuracy that sounds robotic and mechanistic will not lead to the consumer loyalty that Tellme requires. Understanding the need for vocal warmth and empathy, Tellme has done extensive usability testing since December 1999, when it launched a prototype of its voice-activated service, code-named Mini-Me after the Austin Powers character. Consumers, it turns out, are willing to tolerate occasional errors in speech recognition if the service on the other end of the phone is fast, fresh, and friendly. “The one thing people won’t tolerate for very long is the feeling that they’re actually talking to a machine,” says Darby Bailey. An actress by training, Bailey is the voice of Tellme. She’s the one who greets you with “Hey there, welcome to Tellme” when you connect to the service and who helps you navigate Tellme’s top-level menu of choices. While most Tellme applications have their own distinctive vernacular (my favorite is the heavy Scottish brogue of the dealer in Blackjack), Bailey’s girl-next-door voice provides the overall feel and personality of the service.

“The tonality of her voice has the freshness and friendliness we wanted,” Partovi says. “Eventually, we realized that the best place to use Darby is in the home prompt.” That might not seem like a heavy workload, but in a typical week, Bailey, who also does all of Tellme’s radio ads, records roughly 1,000 prompts, ranging from a single word to a few sentences. Often, the prompts are slight variations on the same word or phrase. These subtle distinctions help perpetuate the illusion of a real voice, with all its idiosyncrasies.

This obsession with the nuances of the voice prompt are revealing: Tellme is focused on the user experience. “We want to make sure the service understands the callers and that the callers understand the service,” says Matt Marx, one of the early scientists at SpeechWorks who was recruited by Tellme in September 1999 to head up its voice technologies. The task is daunting; there are a gazillion regional quirks across the country, unlimited opportunities to fall on your face and mispronounce New York City’s Houston Street (that’s “Howston” to you). Marx has 18 full-time employees and 30 to 40 part-time contractors who are busy transcribing the pronunciations of literally hundreds of thousands of words ó restaurants, businesses, cities, streets, sports team names ó that “teach” the speech-recognition engine that Tellme licenses from Nuance. He also works closely with Lisa Stifelman, a Ph.D. from the MIT Media Lab and Tellme’s resident user-interface expert, on the psychology behind the service. “When we built Tellme Sports, for example, we wanted to make sure that the information is always given from the caller’s perspective,” says Stifelman, who played ice hockey at MIT. So if you’re calling from New York and ask for a local score, the service would say, with implicit empathy, “The Yankees lost a tough one to the Indians,” rather than the far harsher “The Indians beat the Yankees.”

Painstaking attention to detail is critical. To offer driving directions, introduced on 1-800-555-TELL last fall, Tellme had its audio team record the names of some 560,000 streets in more than 30,000 cities. The service has been a hit, joining sports, stock quotes, traffic, movies and weather among the most popular of the 17 current Tellme applications

At Tellme’s headquarters, a large U.S. map projected high on the wall shows which applications are being used where. Davis calls this “the pulse of Tellme” and likes to peer out from the sleeping loft over his desk to watch it. Davis, now a seasoned 22, has an unusual role at Tellme. Although he is a cofounder, he disdains conventional hierarchy and has chosen to work for Partovi, overseeing a group of 15 engineers who are writing the code that runs Tellme’s applications.

Partovi and Davis expect to roll out several new consumer applications on 1-800-555-TELL soon, including a much anticipated yellow pages, e-mail, and personal phone book. Tellme will also be offering real estate listings through a deal with Homestore.com. Tellme users interested in buying or renting a home would be able to search by price range and location, and even connect to a local broker to see the listing. “Half the country doesn’t have Internet access and therefore doesn’t have access to Homestore’s listings,” says Homestore CEO Stuart Wolff.

Homestore is paying Tellme to host the application. Eventually, Homestore may also offer the service over its own toll-free number; if that happens, Tellme will host that application as well.

there is some basis to the argument that Tellme is an intuitive idea that, if successful, will be more of a technical than a conceptual triumph. That’s where Tellme chief technology officer John Giannandrea comes in. Giannandrea is a big believer in what he calls first-principle design: taking a very basic approach to solving a complicated problem. Last year, when Tellme decided to restore a 60-year-old British phone booth it had bought on eBay, Giannandrea knew exactly what to do: Pick up a can of authentic “post office red” paint in his hometown of Stirling, Scotland ó where, coincidentally, the phone booth had been made. (The post office originally ran Britain’s phone system.)

Since arriving from Netscape/AOL in April 1999, Giannandrea has been doing the equivalent of creating a language that would allow Shakespeare to chat with hip-hop artist Busta Rhymes ó in other words, building a platform to regulate the seamless movement of voice and data between the existing phone network and the Internet. “What we’re doing, essentially, is bringing Internet technology to answering phone calls,” he explained to me one morning, marker in hand, at one of the many white boards that line the walls at Tellme.

Giannandrea sketched a diagram of the traditional circuit-based phone network on the left side of the board ó lots of scattered little boxes representing phone switches in major cities like New York, San Francisco and Washington, D.C. On the far right he drew a bunch of cylindrical shapes, slightly better organized this time, representing Internet servers. Last, Giannandrea drew Tellme, with its neatly stacked boxes of rack-mounted servers, positioned right in the middle of the white board. Although I wasn’t sure if I had passed his Internet geometry test, Giannandrea had made his point: Tellme was attempting to carve out a valuable market position.

The software that makes this possible is written in VoiceXML, a programming-language standard that is doing for the voice Web what hypertext markup language, or HTML, did for the Internet. VoiceXML allows developers to write voice-activated applications without understanding all the intricacies of speech recognition. As Tellme knows, the more developers who write applications for its platform in VoiceXML, the faster it can become the standard ó a proven technique that has worked for Microsoft and Palm (Nasdaq: PALM) and has hurt Apple (Nasdaq: APPL) because of its closed system. This past June, as part of this strategy, the company launched Tellme Studio, a free Web site providing all the tools, documentation, and resources needed to create voice-enabled phone applications. At year end, 7,500 developers were using it.

Giannandrea also determined early on that if Tellme wanted its network platform to be truly scalable, the startup would need the cooperation of a major phone carrier. This is a major reason McCue wanted John Doerr of Kleiner Perkins Caufield & Byers involved in Tellme, because the legendary venture capitalist had a close relationship with AT&T CEO C. Michael Armstrong.

After Kleiner Perkins invested, in December 1999, McCue didn’t wait long to ask Doerr to make the introduction. “I told John I wanted to talk to Mike Armstrong and do this huge deal with AT&T,” McCue says. “And the very next day, I was on the phone with him.” That Friday, January 28, 2000, conference call is one of the defining moments in Tellme’s short history. On Tellme’s side, McCue and two close associates, Emil Michael and Eric Alexander, were conferenced in on separate lines from their office; Doerr, who was traveling, looped in from his private jet. When Armstrong dialed in from Basking Ridge, New Jersey, he had to get an AT&T operator to break in and ask permission for him to join, because all the reserved lines were already being used. (Everyone had a good chuckle, because the operator had no clue that the Mike Armstrong on the line was that Mike Armstrong.)

That was the call’s only hitch. Its highlight was McCue’s pizza-delivery demo: He gave a voice prompt for an Italian restaurant in Basking Ridge and quickly connected to a local pizzeria that was unaware of its role in telecom history. Armstrong was impressed. Within a few days, Michael and Alexander checked into the AT&T Learning Center hotel in Basking Ridge to hammer out a deal, booking rooms for four nights. They ended up staying two months.

AT&T invested $60 million for a small minority stake in Tellme, but under the terms of the deal no money actually changes hands. Tellme instead gets something more valuable: free network minutes and services, which should greatly shorten the time it takes the company to reach profitability, a critical consideration in today’s unforgiving (if not intolerant) market environment. AT&T also agreed to explore a variety of ways to partner with Tellme, such as comarketing its services to AT&T corporate customers.

Tellme has been even more aggressive in its effort to help AT&T operate more efficiently. The company is close to inking a deal to take over the management of AT&T’s toll-free directory assistance, which receives more than 200 million calls a year and reportedly costs the phone giant upward of $60 million annually to maintain (which it’s required to do by law). If Tellme succeeds in providing this massive voice-activated turnkey service to AT&T, there is virtually no call-center door it won’t be able to open. With more than $100 billion being spent on managing 800-number calls, with “customer service” and “retention marketing” the ringing consumer and B2B buzzwords, the company would be in an enviable position, to say the least. As much of a novelty as it is to ask “What’s the humidity in Miami?” and get an answer, the real revenue for Tellme comes from replacing frustrating, leaden call-center responses with the elegant, winsome chirpiness of Darby Bailey. It also moves Tellme from a straight Internet play to a broader business application that leverages the Internet to replace old-economy inefficiencies ó the model Wall Street hungers for.

About once a month, Mike Mccue holds a special meeting with Tellme’s new employees. At last November’s session, McCue’s easygoing manner quickly put the 25 or so attendees at ease. He kicked off the discussion with his “five-minute elevator pitch” (not surprisingly, it usually takes him twice that long to deliver it). In McCue’s global vision, 1-800-555-TELL would be a self-sufficient ecosystem: Consumers would be able to navigate among phone applications for different businesses without ever having to leave.

Using what’s recently become one of his favorite analogies, McCue compared Tellme’s challenge to climbing Mount Everest and cautioned the group that the journey isn’t a direct shot. “You can’t climb straight up the mountainÖyou have to zigzag.” At the moment, he went on, Tellme is nearing the end of its passage through the treacherous Khumbu Icefall that leads to Camp 1. In his analogy, Camp 1 represents an IPO.

“But what about competition?” asked a new engineer. “Aren’t you worried about AOL and Yahoo?”

McCue acknowledged that both AOL and Yahoo have recently begun offering phone access to users, but he noted that AOL is charging $4.95 a month and that Yahoo still employs a touch-tone system. Other competition could come from startups BeVocal, HeyAnita, and TelSurf Networks ó each of which has introduced toll-free voice-portal services with some of the same features as Tellme, though none offer as complete a service.

But as McCue is well aware, focusing too much on competition can lead any company astray. “The biggest mistake we made at Netscape was to think too much about Microsoft and let that affect what we did,” he told the employees. McCue also told the group that Tellme had to pass one other important test: “I wanted it to be something my mom would understand, want to use and be able to use.”

The sheer ambitiousness and audacity of McCue’s mission is certainly impressive. He hopes to give birth, and voice, to an entirely new industry, an arranged marriage between the low-tech security of the telephone and the muscle, reach, and incalculable data store of the Internet. Will he be able to pull it off, convincing Wall Street, AT&T, corporate America, and his mom that Tellme is the best way to interact with the Internet? Or is Tellme just a transitional technology that will be increasingly less relevant as we all grow more Internet-confident? Given his motivation, it’s hard not to want him to overcome the many obstacles he faces.

Michael Peltz is “Worth’s” financial editor. His feature on Redpoint Ventures appeared in October.