The Secret Life of Bots

Can robots transform customer service in the next decade? Or can they only smooth out the wrinkles? Learn about several cool solutions working today and one killer app for the future.

There were no hell-bent hunks of steel storming the center ring. No crush of metal against blade. No remote-control wizards behind protective glass. Not even an aluminum pulverizer.

Much to my dismay, Boston's recent BOT2001 conference took little inspiration from Comedy Central's geekiest smash hit: BattleBots. A seminar on bots and intelligent agents at work outside the BattleBox arena, BOT2001 spent a great deal more time explaining how the robotic revolution will change customer service than assessing the merits of Ankle Biter versus Vlad the Impaler. However, Boston's bot conference did borrow provocative themes and ideas from Hollywood creations, such as The Matrix and Steven Spielberg's new blockbuster, A.I., that demonstrate the truly powerful potential for bots in the 21st century.

Do I Have to Spell It Out for You?

Many presenters and participants at BOT2001 discussed applications for robot technology on the Web -- particularly for meeting the escalating standards for online customer service. Already, vReps (virtual representatives) designed and deployed by San Francisco's NativeMinds Inc. are answering commonly asked help-desk questions on internal and external sites for Coca-Cola, Ford Motor, and Oracle. By streamlining redundant questions under the bots' domain, NativeMinds says that it's saving its clients money on payroll costs and providing unique information about the user experience.

Among NativeMinds's competitors is Kiwilogic, a European company that also develops natural-language interfaces for the Web. Kiwilogic's Lingubots interact with customers through logical conversation streams. For example, if you log onto a site powered by Kiwilogic and type in "Where is the shipment I ordered last week?", the Lingubot would automatically review your purchasing history and reply with the most recent delivery information.

When a user's request strays from the standard FAQ, the bot hands over the conversation stream to a real, live person for further help. But often, human interaction is all a customer wants in the first place. No matter how intelligent and intuitive bot technology becomes, it will never stem a customer's innate desire to vent loudly and obnoxiously at a human being who is, at least theoretically, paid to make him happy.

Service With a (Simulated) Smile

Companies like LifeFX recognize the importance of personal, one-on-one service and are working to humanize the Web through high-tech graphic interfaces that look, move, and speak like humans. Based in Newton, Massachusetts, LifeFX creates photo-realistic images of people, animals, and imaginary creatures that respond to customers' inquiries with spoken answers, appropriate facial expressions, and even the occasional hoot or holler.

Dubbed Stand-In Virtual People, LifeFX's digital humans greet site visitors, answer customer questions, and appear in Facemail messages. And though the Stand-Ins' words are scripted and their smiles programmed, Steve Ardire, senior vice president of business development and sales for LifeFX, believes that users will begin to regard these synthetic characters as human and form emotional bonds with them, thereby spurring greater site loyalty and customer-service satisfaction.

As cool as it sounds to chat with a digital Bill Gates or Britney Spears, I can't help but wonder whether this technology will really transform customer service on the Web. Will I forget that lost bundle of Christmas gifts if a synthetic Jeff Bezos asks for my forgiveness and sheds digital tears when I demand a double refund? Will I better understand my erroneous cell-phone charges if a digitized pug (yes, LifeFX has created a talking dog) reviews the service contract with me online? Doubtful.

In fact, it's unlikely that any of the customer-service bots in development today will revolutionize the way we use the Web tomorrow. They may provide a quicker answer and reduce operating costs, but they won't change the game -- just the rules. Working with even the best help bot, I am still forced to synthesize, communicate, and wait. I still have to translate my request into simplified terms, and then type it in and figure out whether the scripted answer really solved my problem. I still have to work for a solution.

Dr. Spock Meets Mr. Spock

Perhaps the most enticing and exciting prospect for the future of customer service came from BOT2001 presenter Jack Dunietz, president of Artificial Intelligence NV (Ai), a company that hopes to provide every man, woman, and child with a virtual assistant to live in his or her wristwatch, cell phone, and living-room walls. No logging onto the Web, no dialing 800-numbers, no standing in maddening lines. Just personalized service wherever, whenever you need it.

Want to book airline tickets for an upcoming vacation? Ask your Ai assistant to find and book the best fare. Forgot to buy your wife a birthday gift? Ask your assistant to browse the Web for silver earrings in your price range, and order them for next-day delivery. And what if they don't arrive on time? Don't worry, your assistant is tracking the package and finding alternatives to the gift now lost somewhere over Cleveland.

How is this possible? Well, it's not ... yet.

But Ai has ambitious plans for the future and a compelling technological proposition that just might work. Ai's business plan begins with Hal, a child machine not unlike David -- the "mecha" (mechanical) boy from Spielberg's A.I. But Hal has no body. It's an artificial brain that lives inside a Windows PC.

Every day, neurolinguist Anat Treister-Goren reads children's stories like Hop on Pop and The Little Engine That Could to Hal. A surrogate mommy to the computer baby, she also shares conversations with Hal about cookies, toys, and colors of the rainbow. Like any 15-month-old, Hal responds with some definite answers ("Hal want cookie") and an equal dose of gibberish. But the most intriguing thing about Hal's responses is that they do not originate from some preprogrammed script or language rules. Unlike any bot created before, Hal is not a data processor; Hal is a learning algorithm that acquires language skills through trial and error.

If proven successful, Ai's sophisticated self-learning program could change the way we live and work. Ai expects Hal to converse at adult-human levels within the next 5 to 10 years. By then, Ai scientists aim to replicate and condense the "growing up" process so that Hal's baby brothers and sisters may be introduced to the consumer marketplace quickly and efficiently.

To begin, Hal technology may appear in televisions so that viewers need only say, "I want my MTV" or "Tape The West Wing tonight" to get what they want. Or Hal may pop up in cars, where it could find alternative routes, deliver directions, and adjust the air-conditioning without asking you to lift a finger. The applications for an artificially intelligent, communicative computer are endless. So are the roadblocks -- technological, financial, and social.

The scientific community remains guardedly optimistic about Ai's behavioral model of artificial intelligence, and no one knows for sure whether the company's learning machine will ever progress beyond "Hal play ball." Meanwhile, the private company and the work of its 30 employees is being funded almost entirely by Dunietz, a successful high-tech entrepreneur. And the product remains largely untested by the outside world -- consumers who may trust artificially intelligent products as much as they do bioengineered vegetables.

Hal is not a sure bet by any means, but it is arguably the most aggressive, innovative approach to artificially intelligent customer service around today. And it's just a tad more stimulating than watching Atomic Wedgie annihilate Scrap Metal in the millennial BotBash.

Anni Layne (alayne@fastcompany.com) is the Fast Company senior Web editor. Learn more about Artificial Intelligence NV and BOT2001 on the Web.

Add New Comment

0 Comments