r.
CIO contributing writer Louise Fickel spoke with Lloyd Linnell, senior vice president and CTO at Stream, to find out how Stream's technology infrastructure integrates customer touchpoints to put the right information on its agents' screens.
Linnell has worked in IT management and product development for more than two decades; his resume includes work on early versions of voicemail, ISDN and cell-based network technology. He joined Stream as CIO in 1996.
CIO: Let's talk first about the IT infrastructure that handles all of your calls and e-mail messages.
Linnell: We built a product suite, Emediate, which integrates best-of-breed technologies across all of the various touchpoints: voice, e-mail, chat, collaboration, and Web self-help. We chose Kana as our e-mail engine, Net Effect's NEware for chat and collaboration, and Malloy's Cognitive Processor as our knowledge engine. For our ACDs, we use Aspect in North America and Nortel in Europe. (Automatic Call Distributors route incoming calls, based on the number called and a database of handling instructions.)
We tied all of those products together with IBM's MQ Messaging Bus and then connected everything to our back-end eCRM system. So if a customer were to start an interaction with our company via e-mail and later pick up the phone or do a chat, the agent would have at their disposal all of the past history and know what solution had been undertaken.
You mentioned neural networks. How does that technology work for Stream?
Every time a solution is found to work, the call center agent marks it. Over time the network learns [that] the more a particular piece of knowledge is used, the more likely it will be suggested the next time an agent has to solve a similar problem. The knowledge will age itself out of the system if it doesn't get used for a period of time. It's much like the human body in that if you use something, it gets stronger.
And what about the network itself -- that seems especially relevant as we're talking about making all of this updated knowledge available to your agents. What is your network infrastructure?
We have a frame relay virtual private network that connects all of our sites in North America and Europe. Our frame relay network is augmented in certain areas with point-to-point facilities for very high bandwidth connections. Everything is fully redundant so that if a link or a site goes down, we can re-route it to that same facility a different way. We're moving to an ATM-based network because we're also involved in deploying voice over IP.
What's the significance of voice over IP? What will that technology enable you to do?
Our focus is not on our customers doing voice over IP. It's on our being able to route the call over a private IP network to any available agent anywhere within oour contact centers.
Right now, if I buy a large ACD that can handle 1200 people, put it in a building that holds only 800 people, and then want to open another building down the street, I still have to buy another switch for that other building. If I were to take the link between the telephone and the ACD and convert it to IP -- which is what IP telephony does -- I no longer have any physical distance limitations on where the phone and the switch are located. So I can put a switch in one building and then run all the rest of my telephones over a private IP network and place them down the street -- or potentially anywhere in the world.
Rather than putting ACDs in our new contact centers, our objective is to put them in a hosted environment, such as Exodus [Communications, a data center hosting provider], by Q3 of next year. Then we could host the physical switches along with our key application servers in several locations around the world and route all of our customers' phone and Web traffic into those hosted facilities.
From there, we would build a Virtual Private Network (VPN) that delivers IP traffic to all of our buildings that house agents. And then we could route to the next available agent anywhere within our network. If we wanted to add numerous agents during the Christmas season, we could go to a college campus, rent space in a strip mall for 50 agents, drop in a private line, and run a LAN within the building.
Today you would have to buy a small telephone switch and hire some highly skilled people to deploy it. So it's not economically possible. The client benefits because we'll able to deliver the call to the next available agent even if they're at a different contact center. And that means better quality of service.
We're building a contact center of the future which will be voice over IP with the traffic terminated in a Memphis facility and coming back to Canton, Mass. It will go live in February.
Do you use voice recognition technology in your infrastructure?
We use a limited amount with some of our IVRs (Interactive Voice Response systems) but it's primarily name and number recognition. It doesn't recognize sentences.
Do you plan to move more into voice recognition in the future?
Yes, we're looking at numerous technologies such as real-time speech translation, which involves real-time speech recognition and translation [of languages]. We're also looking at real-time learning, self-help bots, wireless technologies such as WAP (Wireless Application Protocol) and Bluetooth, and remote diagnostics and self-healing for the network.
How do you envision wireless playing a role in integrating customer touchpoints?
One of the areas for wireless is being able to make things like your knowledge base or tracking system trouble tickets available to the person who is most skilled to handle the problem or was handling the problem prior to this contact.
For some of our high-end support where we're working on sophisticated switches and routers, they are more case-based. Those agents could have PDAs or other voice-plus-data devices. As a call or e-mail message comes in, it could be sent using the wireless protocol to the agent -- even if they're not in their physical workspace.
Another interesting area is real-time learning. There are numerous emerging standards that seek to put meta-data around the knowledge content such that a search engine could search for certain kinds of information via the Web.
We're looking at combining that with our quality monitoring. If we monitor an agent during a transaction who is doing poorly in a particular category, we could push training content to them so that they could become better educated in that facet of their job. You could take that same technology and make it Web-enabled for customers.
Your employees provide support for hardware and sofftware companies as well as ISPs. That translates into a lot of different products, many of which are extremely complex and specific (for example, DSL modems from Cisco Systems). How are calls and e-mails routed to the right people?
Most of our agents are dedicated to a particular team focused on a specific company and its products. Routing becomes complex for companies that have numerous products.
To provide better service, you might want agents to cross-train and provide support for five different types of printers and have scores. So the routing gets more complicated. In general, though, it's fairly straightforward.
How is the e-mail routed?
With Kana's sophisticated e-mail routing engine, you map agents and the skills. Then e-mail is routed automatically to the agents, much the same way you route calls with ACDs.
How does Stream evaluate new technologies and then integrate them into its contact centers? For example, when you first started offering e-mail support, how did you decide when the technology was ready for prime time?
It's a complicated algorithm. It's not simply who has the best technology. It's also about the chemistry. We generally like to work with more embryonic companies so that we can put together a consortium of players. If we're talking about real-time learning, for example, we might need to get someone who has an engine to deliver content, another company that can monitor agents, etc. We see which ones are interested in working together. It shakes out quickly to two or three. Then we do a pilot.
When making the decision to go live with a new technology, do you have some sort of performance standard that you need to hit before you flip the switch?
It depends on the application. With voice over IP, we did an internal pilot in August and went live with our first customer in September. We're now planning to expand it from one team of agents to several hundred agents in the first quarter. We'll have a measured approach to deployment. In the first phases, we'll have a fallback position so that if we have an issue, we can support our clients.
Since you've been at Stream, what has been the biggest technology challenge when it comes to integrating customer touchpoints?
Most vendors and software suppliers write their applications with the idea that they're going to sell it to a company who is only going to support their own call centers. In our business, everytime we bring up another client on an e-mail server, we don't want to have to run another physical instance of the software. Because if you're successful and have lots of clients, upgrading software means doing it on hundreds of boxes.
So the biggest issue is the fact that the complexity of our business is far beyond anything that our business partners thought about when they designed their products. We want our software to run in one code base and be able to upgrade and maintain it as one system even though it supports the logical partitioning of a large number of customers.
How do you tackle that challenge?
You have to be very detail-oriented. You have to really understand the details because you win or lose in the details. I've had to read dozens and dozens of technology books and kind of go on my own certification program. Not that I'll ever write the code. But you have to understand it at a granular level to ensure that it will work.
What has been the biggest business challenge?
Probably the transition from being managed by the cost side of the business over to the revenue side of the business where you want to maximize the lifetime value of the customer.
Historically, companies tried to manage down costs. Customer care and tech support were costs of doing business. Yes, they don't want us to talk on the phone for half an hour if you can solve the problem in fifteen minutes.
But if going an extra minute means that customer will walk away very pleased and recommend the company to ten friends, then you're maximizing the lifetime value.
We now quantify things like first-time close, first-time fix rates, customer loyalty, and customer satisfaction. The surveys are now written with lifetime value in mind. We make sure that agents know what their scores are and tie those scores to their bonuses.
Has this transition resulted in a cultural shift in your organization?
You do have to change your focus. It's subtle. It's almost a revival because agents and managers want to treat customers wonderfully. It's inherent in their nature. Now the clients finally understand how important it is to make customers happy.
