AT&T to refine its marketing through artificial intelligence.

JACKSONVILLE, Fla. -AT&T Universal Card Services plans to refine its credit card marketing systems with so-called artificial intelligence.

"Definitely within the next year, we'll be using AI tools," said Leslie Palmer, vice president of emerging technologies at the credit card unit of American Telephone and Telegraph Co.

With artifical intelligence, "you can do it faster, you can do it better, and you know your results are going to be consistent," Ms. Palmer said. "If you can get an improvement, even a couple of points' improvement, that's significant dollars."

AT&T is looking to artificial intelligence systems, which are programmed to mimic human reasoning, as another potential edge in crowded business where it is already a major player.

After three years, AT&T Universal Bank ranks second among MasterCard and Visa issuers, with 16 million credit cards in circulation.

Hunt Pioneered Techniques

David K. Hunt, who recently took charge of the AT&T card business, was widely expected to use advanced technology to build card balances. He pioneered the use of similar techniques as head of Signet Banking Corp.'s card unit.

Experts say any computerized decision-making program could be considered "artificial intelligence," but add that the tools Ms. Palmer is contemplating would allow a marketing department to make subtler distinctions than the credit-scoring models that already are in widespread use.

Various types of advanced programming and networking - neural networks, fuzzy logic, and pattern-recognition programs - allow the computer to say not only "yes" or "no," but also maybe," said Jerome Svigals, an electronic banking consultant in Redwood City, Calif.

The people in the "maybe" category make up "a very important part of the population," he said. They represent a "significant opportunity" for anyone with the tools to identify good risks who might have been rejected under more rigid systems.

"If you're looking for a market segment that would include all 25- to 35-year-olds with an income of $100,000, then that would exclude, in a normal program, 20-year-olds who make 150,000," Ms. Palmer said. "You want to be able to say not strictly |yes' or |no,' but assign grades so you know you're getting all the people who would accept that service."

Ms. Palmer said AT&T would enhance its internally developed software with artificial intelligence products from a growing number of outside vendors.

"If we purchase a neural network, that software would be run on whatever hardware we have," she said. "Then the business people would go through and evaluate the parameters they want to put in their software package."

Mr. Svigals said the move is in keeping with AT&T's reputation as an innovator willing to commit resources to build its business.

More conservative players "like to see the technique being used on 10 million people over a 10-year period" before trying it, the consultant added.

"They could be left in the dust. When these new techniques come to the market, you can't wait."

For reprint and licensing requests for this article, click here.
MORE FROM AMERICAN BANKER