ABN AMRO Supplies Web-Based Help

When Amsterdam-based ABN AMRO needed help managing the online customer service of its U.S. subsidiaries, it engaged Banter, a San Francisco software firm whose suite of applications is based on natural language recognition technology.

Unlike systems relying on keyword recognition, the Banter products use statistics and semantics to recognize phrases, sentences, and ideas in customer inquiries and then generate responses. Based on the Banter Relationship Manager Suite, ABN AMRO's eCARE system will help its U.S. banks-EAB, Standard Federal Bank and LaSalle Bank-to reply more efficiently, effectively and consistently to customer service inquiries that come in through e-mail, chat rooms and, eventually, self-help and co-browsing sessions.

ABN AMRO seeks to provide quality customer service via a number of different channels, including in-person and voice interactions. The addition of the Banter system, which has been running since December, will help to ensure the quality of the bank's online interactions as well.

"We are confident that it will help our customers to have the same extraordinary experience with remote channels," said Maribeth Holback, senior vice president of eCommerce at ABN AMRO, adding that the Banter system is expected to boost customer service effectiveness. "We have found it to be very accurate."

After Banter's natural language recognition component analyzes a customer e-mail or chat message, the system goes into its knowledge base and comes up with possible replies that it "thinks" could be appropriate. A customer service representative reviews these replies and sends the one that is most accurate. A soon-to-be-available self-help feature will provide automated responses in cases where the system has a high degree of confidence.

Banter's Relationship Modeling Engine, its core technology, uses natural language processing to automatically incorporate feedback on the accuracy of responses from every interaction with customers, agents auditing the system via Banter's batch feedback tool, or other customer channels such as e-mail and chat.

In the case of ABN AMRO, Banter reviewed thousands of e-mail messages in order to populate the system with potential questions and replies.

But that's just the beginning. The system gets smarter over time, "observing" as customer service reps respond to questions with which it isn't familiar. When it encounters a similar question later, it knows how to respond. "That's one of the real benefits of the system-it continues to learn," said Holback.

The Banter system's extensive knowledge base actually helps train agents. Observing the answers that the system recommends, they learn about the bank's products and offerings.

"The agents like it," Holback said.

Clients who have used the system-including several banks and other business with high-volume Web traffic-report that its responses are accurate 90% of the time, according to Mark Brewer, vice president of sales at Banter, noting that human agents average about 70%. "They're amazed at how fast it learns, and how it gets smarter."