Speech Recognition

Speech recognition is a lot like IVR, only callers get to speak selections rather than press corresponding numbers on their phone pads to get information.

Speech recognition gives callers without touch tone dialing the same access to information as those with touch tone service. Not only will it satisfy these callers — but think of the population of callers who need glasses to dial. These callers won’t have to juggle their glasses with the phone pad to see the numbers they are pressing.

Although over-the-phone speech recognition still has a limited vocabulary, most systems are effective enough to allow callers to speak selections such as “sales,” “flight number 123,” “transfer cash” or “order baseball cap.”

Speech recognition technology is constantly improving. Vocabularies keep growing (which means you can program the system to understand more caller commands). It seems that almost all systems are now continuous speech.

Make sure you choose one that is indeed continuous speech. Otherwise callers will be forced to pause and wait for a beep after saying every word or number. Since it’s unnatural to speak this way, callers may be more likely to hang up or ask for a rep. There’s also an increased chance of the system not understanding every word, since it’s hard to tell speech from silence.

If you already own an IVR system and want to add speech recognition capabilities, you should check with your vendor. Many of the big manufacturers like Lucent, Syntellect and InterVoice have added speech recognition to their IVR systems.

This technology has made tremendous strides in the last few years. It promises to change the way customers interact with automated systems, broadening the range of telephony interactions.

There are two distinct kinds of speech recognition, known as speaker-dependent and speaker-independent. The two diverge wildly in the kinds of things they are good at, and the kinds of systems needed to make them run.

Call center apps necessarily focus on speaker-independent recognition. Many people will call, obviously. The human brain in the form of a receptionist can recognize a huge number of variations of the same basic input — there are literally an infinite number of ways to intone the word “hello.” What you want in a call center is a system that will respond to the likely inputs — the most common words like yes, no, stop, help, operator, etc., the digits, the letters of the alphabet, and so on.

Internationally, touch tone penetration is still very low, leaving a vast installed base of potential callers who can not access IVR. It follows that these callers are then going to be expensive to process when they come into a call center because they have to be held in queue until there’s an agent ready for them — high telecom charges from the longer than average wait, coupled with the cost of agent-service (rather than self-service).

On the downside, international call centers, particularly those that serve multiple countries, can field calls in multiple languages. If you use an IVR front-end to have the caller select their language then you by definition don’t need speech rec. These are surmountable problems that have more to do with the operation of speech rec in practice than with the underlying technology.

Application Development Software

Getting a voice processing system to do exactly what you want can be frustrating. That’s why call center managers with computer expertise sometimes create their own systems using a PC, voice processing boards and application development software (also called an application development generator, or app gen.

These software packages make putting together a system easier by protecting you from the lower level computer languages (read: “harder to use”) through graphical user interfaces and object-oriented programming.

Sometimes a voice processing system will come bundled with an application development software component, to help you tweak the system to fit your exact needs.

When should you look into using application development software? If you are frequently going to modify your voice processing application — say for each campaign — then you could easily benefit from software that will let you do this yourself instead of waiting for your vendor.

But beware: this is not for the faint of heart. Even the most user-friendly app gen can be a beast. Quite frankly, creating the workflow logic of a voice processing application is very different from knowing how to schedule agents, or manipulate service levels. It’s not always going to work out the way you planned it, and it’s always going to take longer than you think.

Most voice processing systems come with preconstructed apps in their toolkits. If it’s at all possible, try to use those. App gens are wonderful tools, but the closer you get to off-the-shelf, the better off you’ll be.

Voice Mail

Now this is where you get into specialized technology. Not all voice mail systems are alike. And they have not (yet) been completely subsumed into larger boxes (though this is happening at a rapid clip).

A voice mail system answers telephone calls to individual phone numbers or phone system extensions, plays a greeting from the mail box owner and records the callers message.

At the mailbox owner’s prompting it plays back messages, forwards them to other extensions, saves them or deletes them.

Voicemail has a different role in the call center than anywhere else. While it can be used in the traditional sense for call center managers or upper management, it’s most often used to give callers an option to leave a voice mail message as opposed to waiting in queue for an agent when integrated with your ACD.

A voice mail system appropriate for the special needs of a call center should alert agents when there is a message in waiting. If the voice mail system you choose cannot do this, it’s important to create a system designating certain agents to return voice mail calls when call volume falls below a pre-determined level.

Voice mail is a critical tool for the small center that cannot afford to staff agents after-hours. A voice mail system won’t shut out any callers. It keeps your center open 24 hours a day.

Not long ago voice mail was the classic voice processing application and usually came in a standalone system that was sometimes bundled with an automated attendant. Today voice mail is usually a part of a complete voice processing system.

The latest thing in voice mail is “screen-based” voice mail that lets you call up messages of many kinds on your computer screen including your voice mail, email and fax messages. More and more vendors are jumping on the bandwagon to offer computer/telephony interfaces that put voice mail on your desktop PC. With this kind of interface you can get information not only about your voice mail messages, but also view email messages or faxes from your PC. This “unified messaging” gives you one mailbox combining voice, fax and email messages. Unified messaging gives you an easier interface than the telephone keypad. Most unified messaging runs across a LAN and integrates with your phone system. The object is to have desktop control over all of your messages, with the ability to retrieve them, read email messages or listen to voice mail and store and forward them through your computer or phone.

Some benefits:

• Users can view messages that come in even if they are on the phone.

• There’s no need for dedicated voice mail hardware.

• The user can do all the configuring.

• You can move voice, data and email from site to site across networks.

• It eliminates the need to use the phone pad to issue commands.

Previously, telephone-oriented software had to run on systems connected to ISDN or proprietary digital phone lines to control features like hold, transfer and conferencing through TAPI.

Active Voice was one of the first companies to put voice mail on the user’s desktop. Their original TeLANphony product bridged local area networks, telephone systems, voice processing and desktop computing.

With a unified messaging app, when a call comes in, a window on your PC pops up and gives you information about the call. With the click of the mouse you can ask callers to identify themselves, hold, play a greeting, transfer the call to another extension or ask the caller to leave a message without picking up the receiver.

As call centers change (and as the Web becomes more of a customer input channel), combining different types of messaging will be more important, we will probably see the next generation of unified messaging applications focus more on serving the needs of the center.

Automated Attendant

Automated attendants answer a call, play a message with a menu of options and route the caller to the extension or menu choice selected.

In call centers, automated attendants are helpful in having callers direct themselves to an appropriate queue. For example, an automated attendant in a call center might ask the caller to dial one for sales, two for billing or account information and three for technical support. The call would then be routed to the correct department or the correct ACD gate.

Once it was common to find standalone automated attendants. Today they are usually a part of a voice mail or other voice processing system. In fact, they are so fully integrated into today’s PBX systems and messaging technologies that you almost never have to worry about buying this (or any of the voice technologies itemized so far). These are features of larger scale systems that are still important to the functions of a center.

Announcement Systems And Messages On Hold

The most basic building block in the suite of technologies I call voice processing is the announcer. An announcer simply answers an incoming telephone call and plays a recorded message.

Digital announcers use a computer chip to store the recorded message. Other systems use tape to store the message, similar to the way an answering machine does. (Word of advice: stick to digital. Tape is too delicate, too cumbersome, and hard to edit. Digital is not just the future; it’s the present.)

You can have the system play a message and simply hang up, or ring the caller through to your phone system after playing the message if they choose to stay on the line for more information.

Announcers can also work with ACDs to play messages to callers in queue. You can program a message to simply thank the caller for holding, play on-hold music, or even better, play recorded promotional messages.

Because an announcer is so simple, it doesn’t have the high-tech appeal of other voice processing technologies. But announcers are vital to most call centers and many other businesses because they play music and messages to callers waiting on hold or in queue for a call center agent.

Call centers turn to sophisticated technologies like computer integration to save a few seconds per call — and may spend hundreds of thousands of dollars to do so. But few stop to think that a simple announcement on hold that tells callers to have a credit card ready can save that same call center five seconds per call at almost no cost.

When choosing an announcer the most important thing you’ll need to decide is the amount of recording time you’ll need.

Voice Processing Fundamentals

Customers demand convenience. They want information quickly, but they also want specialized attention. And they want to reach you on their timetable, not necessarily yours.

Voice response assures callers reach the right department without the need for an agent. Callers like having options. They hate being forced to wait in queue. Voice processing means you can offer them options. Depending on the technology you use, they can leave a message for a return phone call, retrieve information themselves, or request that it be sent to them.

And the benefits to you are even greater. When you use a voice processing system, more calls get handled through the system. Instead of paying your reps to answer every call, they can handle just the callers who ask to speak to them. Depending on circumstances, you’ll be able to handle higher call volumes with the same number of reps.

Information that an IVR system captures is always accurate. It comes firsthand, from the customer. By now everyone realizes the value of customer information. You can use it for cross marketing, surveying demographics about who your customers are, and so much more.

A lot of information about IVR will be presented in the next section. IVR is a special animal; it’s the key voice processing component in call centers, worthy of special attention. What this chapter will do is explain some of the other voice technologies that are available. These technologies, like speech recognition and automated attendants are, if not critical, then important for specific applications and industries.

Less than 10 years ago it was possible to go through each voice processing technology and give an example of a standalone system that offered that technology. Today’s systems are much more sophisticated.

These days certain technologies are found almost exclusively as functions in larger systems. It’s likely that some or all of these are included in the ACD you’ve already got, whether you use them or not. (Maybe in reading this brief chapter you will see the virtue of simple voice functions and trot out your ACD manual to get some of them turned on.)

When there is a standalone product, it is almost always aimed at the low end of the market. But today’s voice processing market is also a place where you can get what you want — exactly what you want. The hottest technologies are application generation software products, voice boards and the accessories needed to create “do it yourself” voice processing systems.

Here, I’ve outlined the technologies that are available. In a sense, all I’m really doing is showing you how some core voice technology (voice boards plus some software logic) can be put to use. There is precious little difference between auto attendants and their grown-up cousin, IVR, besides power, scalability and feature set. Under the hood, they are all essentially the same. But no matter what type of voice system you choose, they all share one common characteristic — you’ll be more productive and will save money in the long run.

Key Features To Look For | Computer Telephony

And so here are five key features that you might want to add to your call center through computer telephony. There are many others, and the list grows all the time. (That’s part of the wonder of this industry.) Again, these are the ones I see most frequently, that provide demonstrable benefit to the companies that implement them. You can add them by adding software that contains them, or as part of a general CTI overhaul. The benefits of each of them are always the same — faster and better customer service.

1. Simultaneous Screen Transfer. An agent is speaking to a client. The agent has the client’s database up on his screen. The agent needs to send the call to someone else for special treatment. Push a button on his screen, “Who would you like to send this call to?” He types the name, hits Enter. The call and the updated screen go to the specialist.

2. ANI/Caller ID Database Lookup. A call comes in. It carries the calling number. Your ACD grabs the calling number, passes it to your database over your LAN. As the phone rings on an agent’s desk, the agent’s screen pops with a screenful of information on the caller. What he bought last time. What his problems were. How they were resolved. What he tried to buy last time.

   Automatic phone lookup can shave 15 minutes off a typical call — the time the agent takes to ask the hapless caller such questions as “What’s your name, address, phone number, etc.?”

3. Predictive Dialing. When your agents are not answering incoming calls, they could be making outgoing calls. “Last month and the month before, Mr. Smith, you bought four dozen boxes of paperclips. May we send you another four dozen?”

4. Other Database Lookups. Many agents are linked to only one database. But customers always want more information than one database can provide. A price list. A list of your dealers. Other machines you’re compatible with. How to get the machine fixed. Layouts of the hotel rooms you’re renting.

5. Fax server. Your agent is answering a call for help. Explaining the solution is too difficult and time consuming. It’s easier to say, “May I fax you several pages of explanation?” Let your fax server send the pages while you make another phone call. Let your customer follow the jumping ball on the fax he just received.

How to Make it Work | Computer Telephony

There’s no way for me to tell you everything that can go wrong (or right) with a CTI install. There are so many things that can’t be anticipated by outsiders, which is another key reason why you want to have an internally directed plan, rather than handing everything over to a consultant or a systems integrator.

Many companies need help defining the scope of what CTI should do in a business context (not just from a technical point of view). That help can come in the form of a consultant or systems integrator, who typically work with a company to coordinate the entire plan of an installation, help select the products from the various layers, and if necessary create any custom linkages or applications to suit the situation.

Or, that help can come from one of the vendors themselves. This is more common than it used to be, as vendors have worked hard to orient themselves in a way that makes more sense to the end-user: end-to-end coverage of the entire CTI process, from the component layer through the applications and service. They often set up

So use this short list as a tickler, to spark some thoughts about the kinds of things that applies to your particular business circumstances. These aren’t the only things to watch for and pay attention to; they are just the ones I hear about most frequently.

• Single out the high-volume areas of your call center operations. For a telemarketer selling stereo equipment, the customer service and new orders division may field many calls, while the help desk may get a relatively low load. For a PC vendor, however, the help desk may be just as inundated, or more so, in the face of decreased sales.

  Before implementing any computer telephony technology, you should define the internal environment. The areas with high volume are going to have the highest payback when you implement open applications. Possibly, only one or two of your applications would really benefit from computer-assisted telephony.

  The telecom manager should also assess the nature of each department, to see if the switch-to-host application would enhance or besmirch the corporate image. (Yes, failure to do this could prove extremely embarrassing.)

  When calling the complaints division, for instance, the caller usually expects to air her grievances to a live agent. Her resentment and frustration may only build if greeted by a VRU unit.

• LANs, minis or mainframes — size up your host solution. For smaller call centers, a local area network can serve as the entire host side of the solution. Clearly, the last few years of application development have shown that a LAN-based or client/server-based application gives you more flexibility when it comes to importing telephone functions to the workstation than you’d get with a mainframe.

  Of course, if you have a mainframe or mini already in place, you’ll probably work with this existing hardware. You can often use these hosts as central servers, connected to workstations via local area networks, combining the flexibility of a LAN with the processing power of a mainframe.

• Pin down the vendor on probable savings and goals. Before you even think of contacting a vendor, you should evaluate the time it takes to handle a given call. Only then will you know what your projected savings might be.

  Bear in mind that most applications salespeople are just that — salespeople. Get past the vague promises and pin them down on how much will be saved. Demand detailed projections and scenarios. Ask to speak to a few happy customers — even among happy customers you may find some potential drawbacks of a particular system.

  If you’re fortunate enough to be running a regional monopoly — a utility or water works — you can call colleagues from other areas to discuss any open applications they may have implemented. If you run a mail order house, you may have less luck getting your competitors to divulge their secrets.

• Starting over or improving upon the existing order. Many applications can be integrated into an open CTI environment rather painlessly.

  For instance, you may have an application that calls up customer profile information by having the agent key in the customer’s social security number. Using ANI, the open application automatically summons the file to the agent’s screen simply by replacing the social security number with the caller’s home phone number.

  Many open applications, like predictive dialing engines, are more efficient or economical if purchased as turnkey applications. Sometimes, it pays to scrap the old order rather than undergo an ill-fitting adaptation.

• Test the waters. There are two ways to test computer telephony apps prior to full implementation. Dummy applications are available, simulating call traffic, your workforce, the equipment you plan to employ, your network services and your application code. You can also have a test region on the host, where you can run pilot tests while you’re making changes, perform load analysis.

  Many telecom managers prefer to phase in the new regime gradually through such separate testing areas. One way is to phase in with 10% or 20% of your customer base, then gradually broaden the application to include the entire base through the call center.

• Avoid glitz for its own sake. CTI apps perform some feats so stunning that even the most sober telecom center manager can get carried away with the fancy.

  Few would argue that the act of automatically shunting a caller’s vital data and his phone call to an agent’s terminal before that agent even picks up can save a lot of valuable seconds in WATS time and agent labor. All of these precious seconds are lost, however, when the agent picks up the phone and exclaims — “Hello Mr. Brooks, how may I help you?”

  If you call people by name before you give them a chance to introduce themselves, you’re going to waste 20 seconds of your time with ‘how did you know I was calling?’” The result is a transaction three times as long, and three times as expensive, than the manual solution.

• Don’t ignore agent considerations. Weeks before you implement the application, you should set up a training program — coordinated by the applications developer — to master the system.

  As with any introduction of automation, you may need fewer employees on the job — in this case, agents at their terminals. Perhaps your budget will permit you to divert customer service reps to a larger support group or complaint division. You may also be compelled to reduce the workforce, either through attrition or layoffs.

  Also consider the implementation of new evaluation criteria for those who remain at the call center. If your application incorporates a voice response unit, for example, the unit will handle most of the simple inquiries without any live agent intervention. Which means that your agents handle only trickier, more difficult calls. So you should expect that the duration of an average call fielded by an agent will increase.

• Reality checks. Three months after your application is in place, then three months after that, you should take a look at how much you are saving.

  Note that while it is relatively easy to calculate lower toll free usage or fewer agents needed to staff the phones, other benefits are more difficult to gauge — such as how many new policies have been taken out by insurance customers simply because the agent was able to transfer both their datafile and screen immediately from the life insurance division to the accident group.

  Often, you’ll find you must alter your long-distance contract, your agent scheduling, even the capacity of your computer plant to accommodate the changed call processing environment.

  From a bottom line standpoint, though, these changes are probably for the better. Many end users report a nine to 16 month payback on their investment.

The Call Center Jigsaw Puzzle

Putting the pieces of a CTI system together involves an amazing degree of coordination between products and vendors at several levels.

The bottom layer consists of the fundamental hardware and conjunctive elements: the boards that process the voice and data channels; the servers and networks, often ruggedized to reflect the mission criticality of what they are used for; and the standards and open APIs that link different vendors’ equipment together. The most common boards used in CTI systems are from manufacturers like Dialogic, Natural Microsystems, Lucent, Brooktrout and several other specialty companies, depending on the application.

Parallel to that sits the dual networking infrastructures: the phone switches and the data networks. The phone switches are usually PBXs or dedicated high-volume call routing switches called “automatic call distributors,” or ACDs. Phone service is also a core component. Not just because it’s an obvious necessity, but because increasingly, the carrier networks are being upgraded to deliver advanced call processing services through the network. Sometimes this works directly to the advantage of the smaller business — if messaging or call routing applications can be run from the network, you need to invest less in premise-based equipment. You can implement “high touch” services like call centers without spending so much on high tech infrastructure.

The data networking infrastructure, like the phone system, is probably already in place: LANs, intranets, external Internet connections and websites, desktop browsers and firewalls.

Between these two networking areas lies the middleware layer. The products in this category are what most people think of when they say CTI — the very specialized applications that draw data out of host systems and coordinate it with incoming telephony information, then format it for both sides. Originally, many of these products focused on coordinating between a single vendor’s switch and a single host format. As a rule of thumb, the older and more widespread the databases, the more important (and more complex, and customized) the middleware has to be. This has accounted for a lot of the tension surrounding the installation of CTI. For companies with decades-old legacy systems and extremely customized databases, installing CTI meant that to achieve any of the benefits, you had to go through a trying period of intimate customization between the switch and the database.

Increasingly, middleware connectivity is being sold as part of the switch, and middleware companies themselves are being acquired by larger companies above and below them on the CTI component chain.

The next level of product in the CTI hierarchy is the application layer. This is the software that actually does the things that make people more productive, things like messaging or speech recognition, automating sales forces or taking orders through the Web. It’s a good idea, when pondering a transition to CTI, to start here, with a concrete idea of what you want the system to accomplish. It’s akin to buying a PC based on what kind of application you want to run. You pick out the spreadsheet and word processor that has the features you need, then buy a PC that makes those features work best. CTI is no different. The best approach is to identify the applications that suit your business and then build up and down to integrate those apps with the infrastructure you already have.

Along with these layers of technology come the consulting services and systems integration know-how that ties it all together. For the most part, CTI is not an off-the-shelf accomplishment. It does require intimate connections between different technical realms — which are usually managed by different people, with different sets of priorities.

What it’s Used For | Computer Telephony

Voice response systems are front-ends to the phone system that deliver recorded information when someone calls. Interactive voice response is two-way; it responds with information when a caller enters digits on the touch tone phone. And when that information comes from a host database, that’s CTI in action.

Customers can call at any hour of the day or night looking for account balances or order status information. The IVR engine queries a database in the background and reads the information to the caller. In this way it can be made dynamic — instead of just reading off a set of canned, pre-recorded announcements (“we’re closed right now, please call again tomorrow” or “to leave a message for our sales department, press one”) it pulls real time data out of corporate databases.

When this is translated to the Web, the kind of information you can make available to customers is expanded dramatically. Anything visual, from catalogs to product schematics, can be dropped onto a customer (or agent) desktop. Customers can help themselves when problems arise. They can learn about your products before they buy. And when it comes time to talk to an agent, they are better prepared; so the call is shorter, more effective, and more profitable. The “shopper” does his shopping without consuming your most valuable resources. But the buyer gets your full attention.

CTI is an information delivery tool. It won’t make a seller out of someone with no sales ability, but it will give someone who deals with customers the knowledge they need to address the needs of the customer.

For starters, a customer record is brought to the agent’s desktop at the same time as the customer’s voice arrives on the phone. The caller can be identified in many ways, including the information that travels with the toll free call, or through digits entered by the caller himself. When the agent has the customer information in front of him, the call doesn’t last as long. The customer doesn’t have to repeat himself every time the call is transferred. And the agent sees the entire history of the relationship with that customer. If the customer has a history of problems, the rep will know about it. And if the customer has a million-dollar lifetime value to the company, the rep will know that too.

Better still, if that caller is really a million dollar value, the CTI system will know it before the rep will, and can be set up to send the call directly to someone equipped with the experience to handle priority customers.

Another way CTI helps is with quality control. In call centers, calls are often monitored — recorded and archived so that the agent and his or her supervisor can listen to them later and assess performance.

That analysis process is made much more productive when it’s augmented by the data that passes through the agent’s screen during the call. A complete record of every transaction can be kept indefinitely, including every screen viewed by the agent. This “screen scrape” is an audit trail and a training tool.

All these things help a company cut operating costs by reducing (or stabilizing) support staff headcount. A company can be more productive with the same staff by handling more calls (or customers, or transactions, depending on which metric is most important to them).

And most important, it lets a smaller company look like a big one — without sacrificing the personal touch. Customers don’t care how big a company is, they care about what kind of response they get when they call or contact that company. If they get good service from Federal Express or L.L. Bean, they’re going to learn to expect it from every other business they deal with, no matter how large. CTI applications allow companies to appear more fully outfitted than they really are. This could mean putting systems in place to answer calls during off-hours when no agents are available, or having a website take orders at all hours. In any of these cases, the underlying technology that links the computer networks and the telecom systems enables the small company to decide for itself how to manage its customer relationships.

A call center that uses computer telephony knows who its customers are and why they are calling. It knows what they like, what they dislike, and how much they are worth to the company. On the other hand, without computer telephony every customer interaction is like a blind date — full of expectation, and possibly, frustration.

CTI lets a company respond faster to changing market conditions. But it must be implemented correctly: with clear and ongoing support from upper management and a clear-eyed view of the company’s goals for the technology.

For a company to put computer telephony into place requires that they determine, from end to end, exactly what they want a customer interaction to be like. Every contingency must be accounted for: phone calls; emails; fax requests; even Web hits. Far too many companies have had disappointing results because they didn’t put in the computer telephony they needed; they put in the technology they imagined. The right CTI is the mix of applications and core technologies that add value to the company’s existing operations, and allow it to do more: voice mail, unified messaging, advanced call routing, fax redirection, Internet telephony, call center apps, customer service software, sales force automation — whatever combination is most useful in their particular circumstances. The key is to figure out which pieces are right for which circumstances.

There are many ways to make it work. There are also many places to go for expert advice. Component vendors start the process, and often point the way to application partners whose product works with the core pieces. Telcos and other large service providers can also provide an umbrella under which integration between all the pieces are certified to work correctly. There are systems integrators, who specialize in matching the various pieces to the custom needs of a particular company, or vertical industry.

Whichever direction, the growing company needs upper management buy-in, direction on the goals of the project, and a clear-eyed view of the relationship between the company and its customers.