After years in the customer support business, Michael Doyle has dealt with just about every kind of help desk emergency you can think of. Earlier this year, however, Doyle faced a situation that finally pushed him to the edge: His company, Logitech Inc., was about to launch a new product—a $49 camera for sending images over the Web—and Doyle expected a flood of customer calls and e-mail would overwhelm him and his help desk.
But Doyle, ever the pro, didnt panic. Instead he hired Lily, a Web-based automated customer support bot that never needs a lunch break or a day off. With Lily working round-the-clock, Doyle and his team managed to handle all the customer queries and save Logitech a pile of money at the same time.
At a time when e-commerce profits are down, margins are thin and the number of new, inexperienced Web users with lots of questions is growing, e-businesses such as Logitech are facing a customer support dilemma. The traditional solution—a call center manned by trained customer support reps—can cost $20 per call, or more, according to some estimates. Moreover, human staffers arent easily scalable—for new product launches, for example—and they dont work nonstop.
Enter a new generation of smart, self-learning Web products such as Lily—hosted service bot software from NoHold Inc.—that can answer e-mail personally, converse with users and even project a personality, of sorts.
Such technology can drastically cut the cost of providing online customer service—to around 25 to 50 cents per call—while helping e-businesses gather information about exactly which questions are being asked, how often and how theyre resolved.
But, experts say, you need to pick the right bot for the job. A company selling directly to consumers, like Logitech, might want a chatterbot with a persona, while a business-to-business site might simply need to get the job done efficiently.
So, its important to examine the underlying technology before choosing a service bot product, experts say, and to resist the temptation to simply give consumers online access to customer service automation systems originally designed for help desk professionals.
Theres little doubt that most e-businesses still have a long way to go to get online customer service right. Some experts say 50 percent of Fortune 1000 sites lack any kind of Web-based help. And thats a huge mistake, said Joe OLeary, managing partner for Arthur Andersen LLCs worldwide customer relationship management group in Chicago.
"Every time you have someone come to your Web site and look for help, you have an opportunity," OLeary said. "If you arent prepared to take advantage of it—and most companies arent—youre missing a big opportunity."
But, apparently, many e-businesses intend to correct that problem soon. According to a recent survey by Gartner Inc., in Stamford, Conn., e-businesses see improving customer service as their No. 1 spending priority in the coming year.
Automated customer service bots represent one practical way to beef up online support without going broke. Web-based customer support products come in a variety of flavors. Products that automate e-mail handling typically scan the content of the users message and, acting on certain key words or by synthesizing the content, respond appropriately. Live chat or help products such as Lily operate by dissecting the user request and then generating the appropriate information, all in a chat-room format.
Best-of-breed products have the capability to understand the message and the context, ask the user for clarification, and respond. When the question is asked again by another customer, these systems "remember" what theyve said and can answer more quickly.
As powerful as that sounds, e-businesses shouldnt start sending pink slips to all their human help desk agents. Thats because, experts say, bots arent right for all customers and all queries.
"[Vendors] are claiming call resolution rates of 70 percent to 80 percent [using bots]," said Mitchell Nitzan, an analyst with Gartner. "Were seeing rates a lot closer to 50 percent. Id check every reference I could find."
While youre at it, check the technology underlying the various types of customer service bots on the market. Most of the more sophisticated bots—the ones that converse in natural language and learn about customers from the questions they ask—claim to be based on artificial intelligence engines. That may be true, Nitzan said, but some products based on AI—especially those that are rule-based—can be so kludgy as to be useless.
Live or Lily?
Before finding Lily, Logitechs Doyle, in Fremont, Calif., was well- aware that picking the wrong bot could actually undermine customer confidence. "The worst-case scenario is where the system ends up dumber than the customer," he said. For him, the make or break was whether a product could actually engage in a dialog with the customer. "If theres a back and forth, you end up with a much better result, and its much more like talking to a live person. With the so-called natural language query systems, like Ask Jeeves, theres no clarification and no back and forth. Thats just not how a technician works."
Lily is actually the persona for the DynamicDialog product developed by NoHold, of Milpitas, Calif. (Shell answer to any name you decide to give her.)
DynamicDialog looks like a chat room, and it asks users interested in the companys $49 camera to post questions in natural language. Using an AI engine, Lily then answers them, or asks for more information, and can even respond to a change in subject. And Lily can remember answers to similar questions so theyre readily available again.
Its an interactive process that, Doyle said, Logitechs customers find comfortable and natural.
To date, Lily, which is hosted by NoHold, has made it possible for Doyle to minimize the need to add to his staff of help desk agents despite soaring online customer service requests. And, he said, the technology was relatively inexpensive and easy to deploy. To get Lily to blossom, all Logitech had to do was turn over the relevant manuals, FAQs and other data to NoHold, which put that information into the database. In about a month, the product was up and running.
Now, when Doyle and his team look at the statistics in NoHolds Knowledge Portal (the information manager), they can see which questions customers are asking most often and which arent being answered. From the portal, Logitech officials can also e-mail NoHold with new data directly to add into the database.
Its still in its early stages, Doyle said, but he hopes to roll Lily out to the rest of Logitechs product line over time.
Logitechs smooth deployment notwithstanding, experts caution, setting up an automated Web-based customer support system is not necessarily painless—even if its a third-party hosted application.
Transferring the knowledge currently residing in the heads of customer service representatives is no mean feat, and its almost as hard as tapping into a variety of back-end systems for other data that could be needed by a bot. Collecting the large amount of information you may want to make available to customers through the bot can be daunting, experts say.
"If its spring, and someone is visiting the John Deere site, it makes sense to offer information about buying a lawn mower, no matter what theyre looking for," said Greg Gianforte, CEO of RightNow Technologies Inc., a Bozeman, Mont., maker of Web-based customer support tools. "And if its summer, it should be easy to find information about repairing a lawn mower, etc. Thats where a lot of the thinking has to come in. How can we help and give the user access to more of what we have to offer?"
For some e-businesses, a chatterbot like Lily dripping with personality is not necessary to keep up with online service requests. At Resume.com Solutions Inc., an automated e-mail response system from Banter Inc., of San Francisco, was enough to stem what could have been an overwhelming tide of customer questions.
From December to the end of January, business doubled at the résumé-writing site. With just 13 staffers—four were handling customer support—it was getting impossible for the small company to keep up with all the e-mail.
The solution came in the form of Banter Reply, a personalized e-mail program that gets back to customers in a matter of minutes with answers to their queries. The product scans the e-mail for key words, then crafts a response tailored to those words. And it gets smarter and faster as it goes along. The Banter Reply program has been up and running at Resume.com for a few months now.
"Business doubled, and we didnt skip a beat," said John Waller, CEO of the New York-based company. Resume.com was able to scale back to one customer support rep, and this person only has to do support part time.
"E-mail is like a black hole," Waller said. "The companies that are going to win are the companies that actually take care of it. No one can grow costs, but you cant afford to ignore this, either. You just have to take care of it."
Hes so sold on this philosophy that hes starting to deploy Banters self-service bot, which includes online chat.
Another e-business that has gotten good results out of a smart, automated e-mail response system is Rockford Corp.s Rockford Fosgate brand auto electronics group, based in Tempe, Ariz.
Faced with a rapidly rising number of e-mail information requests from online customers, CIO David Richards deployed RightNows Smart Assistant in 1998 rather than increase the number of hours his help desk was available. Since then, the company has successfully solved more than 93,000 incoming queries.
"Its a pretty smart product," Richards said. "It scans the e-mail for key words and then sends back a personalized thank-you note, along with a list of related FAQs which could help. And obviously they do, because for a lot of our customers, this is plenty of information."
The benefits, he said, are clear: far fewer phone calls—and those that do come in are sent to the Web site if they havent already been there—increased coverage and fewer staffers. Before installing RightNow, Rockford had two full-time techies answering e-mails and a number of part-timers. Now one person handles e-mail and phone support. Its worked so well the company is working on adding another RightNow module to allow for live chat. "We used to receive 50 percent to 60 percent of our calls on the weekends, and those people had to wait. Now were helping people 24-by-7. Its really working," Richards said.
Making them drink
In some cases, automating online customer service is only part of the battle. Youve also got to come up with creative ways to get customers to use the Web site rather than pick up the phone and call the help desk.
Just ask Kent Barth, senior director for customer value systems and e-business development for automotive systems manufacturer Arvin Meritor Inc., in Troy, Mich.
About a year ago, Barth and his team created a Web site for the companys Light Vehicle Aftermarket division. The goal was pretty simple: reduce calls to the 30-person help center by creating an easy-to-use Web site where customers could get help, ask about products and check on order status.
Soon after launching, Barth and his team did an analysis of where hits were coming from and what impact they were having on help desk call volumes. What they discovered shocked them. While the new Web site was pulling in 8,500 hits per month, one business customer was still using the help desk to the tune of an average of 600 calls per day. It was clear that simply launching Web site help alone wasnt going to solve this problem.
So, after meeting with the customers IT team, they decided to port a piece of the Arvin site directly onto the offending customers intranet. The project is still in the works, but Barth said he thinks it was the right decision.
"Weve really validated what we were setting out to do. Weve set up specific areas of the site to help people through their questions, and if we cant answer them online, to get them to the right person quickly," Barth said.
In addition to using automated online service software to cut down help desk calls from consumers, some organizations, such as DaimlerChrysler Corp., in Auburn Hills, Mich., are deploying online systems for business partners.
The vehicle manufacturer maintains a help desk that takes calls from dealership mechanics seeking help trouble-shooting car problems, including everything from front-end shimmies to loose lug nuts.
Recently, however, as call volumes increased, Priscilla Hedin, the companys service technology resource center manager, started looking for ways to drive more help queries to the Web and to automate the answers. At the same time, Hedin didnt want to compromise on her 98 percent call completion statistic or under-2-minute hold-time goal. But she didnt want to add staff, either.
So Hedin did what some experts say not to do: She rolled out a tool originally designed for her help desk staff to mechanics in DaimlerChryslers 50 dealerships around the country on a pilot basis. The mechanics access the system, based on eGain Communications Corp.s Knowledge Agent product, via the Web.
How has it gone? Although experts warn that repurposing a help desk tool for a different audience can create confusion, the test drive has gone so well that the product will be rolled out to all DaimlerChrysler dealerships earlier than planned, likely by the end of the year.
Hedin said the product is so easy to use and thinks so much like a mechanic would that it was an instant hit at the dealerships.
This works for DaimlerChrysler, said John Ragsdale, an analyst with Giga Information Group Inc., in San Jose, Calif., probably because the information in the database doesnt change constantly.
"I dont think products for the help desk should go to end users, but at least with a car company, youre not constantly having to add new models and update the database with new information," Ragsdale said.
For most business-to-consumer applications, however, technology that can at least simulate a bit more personality would be best, experts say.
And thats where a persona like Logitechs Lily comes in. She may not be capable of replacing all your human help desk agents just yet. But at least shes more than willing to work right through the lunch break without a word of complaint.