×

TMCnet
ITEXPO begins in:   New Coverage :  Asterisk  |  Fax Software  |  SIP Phones  |  Small Cells
 

Betsy, an inbound contact center agent, handles about 100 calls a day. She is always polite and courteous, except for perhaps late in the day when she gets tired of reading the same script repeatedly. She gets two weeks' paid vacation, 10 paid holidays, and a half hour for lunch every day. On average, she calls in sick five days a year, is late to work 30 times, takes a 15 minute break every four hours and goes to the rest room two times per shift. She gets paid $15 an hour and requires health insurance for herself and her family. She asks for a raise about every six months.

Gladys, on the other hand, can handle thousands of calls a day — hundreds simultaneously. She is polite and courteous and never gets tired of reading the same script. She never takes a vacation, doesn't celebrate holidays, never eats lunch and never calls in sick. She works 24 hour shifts and never takes a break. She costs about 20 cents per transaction and requires no health insurance.

So is Betsy jealous of Gladys? Of course not …Gladys is a machine. She is an automated agent — a software solution that uses speech recognition to carry out "natural" interactions with people over the phone. If anything, Betsy likes Gladys because she makes her job easier.

Gladys can help customers with basic information and transactions. She can tell a customer when his order was shipped, what items he ordered and how much his credit card was charged. She can help the customer find a particular product in the catalog or on the Web site, tell him what sizes and colors it comes in, arrange for shipping, take the customer's credit card information and complete the transaction. Betsy and her fellow human agents are no longer bogged down with routine transactions and can instead concentrate on handling more complex interactions which have a higher transactional value. Thus, a contact center can blend automated agents with live agents to whatever degree of efficacy it desires, automating those interactions which are suitable for automation, while leaving more important transactions for the live agents.
When a customer needs something that Gladys can't deliver, or says something that she doesn't understand, she automatically transfers the call to Betsy. And when Betsy gets the call, she has all of the customer's call history right on her desktop, including information on their past transactions, plus everything Gladys told the customer and what offers she has already made. With this detailed information, Betsy can quickly pick up where Gladys left off and help the customer get what he needs. And if at the end of the call there's a need to complete another transaction, Betsy can hand the call back to Gladys, who is happy to finish the sale, achieving the much sought-after first-call resolution.

Welcome to the world of contact center automation, where human and machine work in harmony to provide superior customer experience. Playing an increasingly important role in this automated world is speech recognition and analytics. Speech recognition is what Gladys uses to interpret what customers are saying so she can carry out their commands and give them the proper information. Meanwhile, speech analytics is what contact centers use to "mine" recorded calls to detect problems with agent/customer interactions and to gain insight into customer behavior which the enterprise can then use to drive key business decisions. Speech recognition and speech analytics are two distinct yet similar disciplines, each yielding benefits for the contact center in the form of operational efficiencies, higher customer satisfaction and increased profits. Let's take a closer look at the ways speech technology is being leveraged in the contact center today.

Speech recognition has advanced in recent years to the point where today's automated agents are able to ask callers open-ended questions, accurately interpret their responses and determine why they're calling. This achievement springs not only from significant advancements in the core technology, but also from innovations in application design, user interface design and the speech science capabilities used to optimize the software for specific user demographics. Speech technology vendors have also made strides in recent years in terms of technical optimizations, and through experience, speech software vendors and their customers have gained a better understanding of where it makes sense to use speech applications and where it doesn't.

Natural Language Capabilities

Their ability to carry out more natural conversations with callers is perhaps the most impressive aspect of today's speech recognition solutions. Thanks to the advances in large vocabulary continuous speech recognition (LVCSR) systems, automated agents can now ask open ended questions such as "How may I help you today?" and either provide the appropriate response or, in the event the system doesn't understand the caller or can't help, steer the call toward the appropriate agent. When implemented properly, these speech self-service solutions can lead to higher customer satisfaction, as the customer is no longer limited to a static set of options, one of which must be selected using the telephone keypad in order to proceed through the call.

In addition, because they are becoming so proficient at interpreting what the caller is saying, today's speech recognition solutions are able to serve as the ultimate automated "switchboard operators," in that they can intelligently route calls to anywhere in the organization, based on the words and phrases a caller uses. As such, organizations are speech applications to unify all their numbers (such as numbers for separate departments, divisions or subsidiaries) under a single toll-free number which can be used as a starting place for all incoming calls. And unlike a live switchboard operator, the system can handle sudden spikes in call volume, potentially handling hundreds or even thousands of calls coming in at once.

Daniel Faulkner, senior solutions marketing manager at Nuance Communications, said today's automated agents deliver tremendous value to the enterprise, just through their advanced call routing capabilities. "There's an awful lot of error in human-based routing," he explained. "And for the customer, it can result in some pretty horrible experiences."

An example of this, Faulkner said, is the typical scenario where a customer gets transferred to the wrong department, but then the agent in that department cannot transfer the call back, so the customer must hang up and dial back into the system. By centralizing all of an organization's numbers, the caller can always get back to the main system, where the automated agent can then re-direct the call.

Faulkner said another big advantage of today's automated agents is that they let the customer decide how he or she wants the interaction to go. This is particularly helpful in situations where customers use a variety of terms to describe what they're looking for. For example, someone calling a tech support center "might call in and explain the problem in extremely technical terms — whereas someone else might use very non-technical terms." Either way, the speech engine can accurately interpret what is being said and route the call intelligently.

"We've sunk a considerable investment in an area of technology called conversation management, or dialog management," Faulkner said. "And that's a really exciting technology because it makes our solutions much more robust to certain types of errors. When people ask related questions that an older, more rigid speech solution couldn't handle, we're finding that we can now handle those. We can cope with a lot more of those areas where people sort of go off topic … we're finding that we can handle that non-sequitur question, and have them continue through the system, rather than pass them off to an agent."

Faulkner added that Nuance's solution has evolved to the point where it can either let the customer guide the interaction, or it will "drive" the conversation and the customer can sit back and enjoy the ride. "What we really want to do is enable people to carry out the transaction how they want to carry it out," he said. "Our goal is to make our solution so robust that people can carry out the type of interaction they want."

Rapid Configurability
On the flip side of letting the customer drive the interaction is the new trend of giving the enterprise greater control over the way the speech application works.

According to Hollis Chin, marketing director for Voxify, his company's new release was designed to give the contact center "full control to make real-time changes to speech applications."
"We have a vision of what we're calling 'consolization,'" Chin explained, "which is essentially putting Web consoles into the hands of the contact center managers — who want the control, the self-sufficiency and the immediacy of making their own changes to the speech applications."

Roger Nunn, Voxify's VP of sales, said providing this level of flexibility not only gives the organization the ability to adjust call routing on-the-fly, it also enables companies to quickly add or take away information about their offerings, and change the way they market their products. "Your user interface says a lot about your brand," Nunn said, "I've read some articles which talk about self-service as being a continuation of your differentiator. So these things are no longer necessarily just contact center-centric. They can be an extension of other types of self-service you're offering to your clients. Most of us think of the Web interfaces as self-service interfaces — but you need to offer similar features and capabilities through your speech-based self-service as well," he said. "Sure, it's your brand, your message — but it's also about the ability to change your offer. The example I give is: nobody today will have the same Web site six months from now. Those are living breathing things which respond very dynamically to the market, to the customer, to the products that are being offered and so on. And we believe that speech-based self-service needs to do the same thing: the old boring IVR that is the same as it was two years ago isn't going to hack it."

Secure Transactions
As Nunn explained, part of the beauty of automated agents is that they can handle transactions in a highly secure way — in that a live agent never needs to be involved in the transaction. "We get a lot of customers coming to us today to talk about PCI (payment card industry) security standards," he said. "This is the data security piece — this question of whether you are securing your customers’ credit card information. It's pretty well known now that the biggest source of fraud for credit cards in a phone-based ordering system is the agent taking the credit card number. Because what they do is write down your information on a piece of sticky paper, and they know everything about you, they've got your name, your card number, your mother's maiden name, that funky three-digit number on the back of the card, your mailing address. And an amazing number of them come into these centers, take down the information, walk out the front door and commit fraud."

With today's automated agents, you can have the system take all of the credit card information and process the interaction in a secure way. Nunn said this gives customers greater peace of mind and increases the value proposition for speech self-service systems. "Even if you want your agents to take the order, why not automate the back-end piece?" he asked. "That way, as soon as the customer says 'I'm done. I don't want to order anything else,' you can send them over to an automated system that takes care of the transaction in a secure way that it not recorded. So you get loss avoidance, as well as cost avoidance. And customer satisfaction is increased, as no human is involved in the actual transaction."

Automate Outbound Campaigns
Another area where speech technology is creating tremendous efficiencies in the contact center is its application in outbound systems. Today, it is not at all unusual for centers to conduct fully automated outbound sales and marketing campaigns, using a combination of predictive dialers, IVR systems and automated agents. This is where a customer receives a call at home, listens to a recorded message and then interacts with an automated agent who first determines the caller's interest in the product or service being offered, handles any questions the customer might have and processes the transaction. In the event the system doesn't understand or cannot help the customer, the call is typically then transferred to a live agent.

"With these new outbound systems, you can automate the entire call," Nunn said. "For example, if I was a pharmacy company, and I knew that your prescription had expired, or was about to expire, I have two choices: I can wait until you walk back into the pharmacy and renew it, or I can outbound call you. Now in the old days I'd have to manually call you and leave a message and say 'this a courtesy call from so-and-so and I just wanted to remind you that your prescription needs to be refilled, and you can call us back at this number and re-order it.' But why wouldn't you go one step further and have the system ask you whether or not you want to renew it? And if you do, the system can ask the customer whether they want to pick it up in the same location or have you mail it to them. And if it's going to be mailed to you, 'What's the address where I should mail it? And by the way, I see I have a credit card on file.' So I never needed a live person — we see that as the next step."

Co-worker Or Replacement?
Obviously the main thrust behind adoption and implementation of speech-based self-service solutions is to reduce headcount and save on salary costs. On average, staffing represents anywhere from 70 to 90 percent of a contact center's operating budget, so it makes sense that companies are looking for ways to automate processes.

"Yes, we do have numerous customers who have reduced contact center staffing through the efficiencies gained by implementing speech solutions," Faulkner said. "There is an example of that — Lloyds PLC, a bank in the UK, actually shut down one of its outsourced contact centers because they were having so much success with their speech recognition solution in the UK." Although reducing headcount is one of the main selling points for speech self-service solutions, most companies aren't looking to publicize layoffs. Furthermore, as Faulkner explained, reducing the headcount "isn't always the end game for the enterprise." Sometimes organizations do it purely for the purpose of providing improved customer service.

"A lot of our customers today are prioritizing the customer experience over efficiencies," Faulkner said, adding that many of Nuance's customers "have us hold back on the efficiencies for the purpose of pleasing the customer." He said even though a call can be completely automated and doesn't have to include a live agent, many companies see a high value in putting a live agent on the line at some point during the interaction. So in most cases, contact centers are using a mix of automation and live agents for the purpose of reducing agent call time.

"We can provide automation rates that are down around 20 percent, all the way up to nearly 100 percent [of the interaction], depending on what task is being automated," he said. "We tend to automate tasks that a) happen a lot — you really only want to invest automation if you're going to automate something that your customer frequently like to do; b) that lend themselves to automation; and c) that deliver a strong ROI — in other words, that it is of a duration or complexity that makes it easy to automate."

"If I'm going to offer a self-service capability, the types of transactions I want to automate are the ones where there are enough transactions that it is worth the trouble," added Voxify's Nunn. "And the example I give is, I could build a Russian self-service application, but if I only get four calls a year in my Alabama call center, is it really worth the money I'm going to spend to do it? So you want to only automate the things that are worth automating."

Nunn pointed out that some transactional calls are very cost-effective to automate. "If I'm going to buy a diamond, I might want to talk to the person live, and they're going to tell me a lot about it," he said. "And the transactional value is so high that it is worth customer touch. But suppose my average transaction value is about a buck. I'm going to want to automate that all the time, because it pays me to invest in that sort of thing."

Part Two of this article will be presented in the January 2008 issue of Customer Interaction Solutions.

The author can be reached at [email protected].

CIS Table of Contents
 
| More