CA BUILDS ARTIFICIAL INTELLIGENCE INTO SOFTWARE
by Wylie Wong

It sounds like science fiction: Software that allows computers to seem more human, exhibiting the ability to think and predict future events. But how close is it to becoming reality?

For more than a year, Computer Associates International (CA) has hawked its vision of artificial intelligence software that can analyze information, forecast the future and help companies plot their computer and network management strategies. The business software maker initially sold the technology, dubbed "Neugent," as a way to monitor the health of computer networks.

Neugents are small pieces of software code distributed throughout a network. They collect data from network equipment and computers and can warn businesses of potential problems, such as impending outages.

The company released new Neugent technology that takes the software one step further by helping information technology managers formulate a computer resource strategies.

Use of the technology underscores the subtle nuances of the lucrative network management software market. Companies such as CA continually look to beef up their management software tools to gain an edge over competitors.

Analysts believe CA is the first software maker to try to build artificial intelligence into mainstream technology to help companies run their businesses. It's unclear whether CA's rivals plan to enter the market. Hewlett-Packard and IBM subsidiary Tivoli Systems have been mum on their plans, while BMC Software is testing a similar technology in its labs, said Illuminata analyst Jonathan Eunice.

Only time will tell how effective CA's Neugent technology is, according to analysts.

"There was some question a year ago about how able this technology was, but I've been impressed," said Hurwitz Group analyst Richard Ptak. "CA has spent a considerable amount of effort to make these things learn, collect and correlate data and be able to make deductions based on information they have."

But the proof will come from businesses that build Neugents into their computer systems, Ptak said. "CA demonstrated the ability to use it, but now it has to be proven. There's not a lot of wide body of evidence that it's able to deliver."

CA executives say more than a hundred customers are using Neugents to monitor the health of their computer systems, and about 2,000 others are testing the technology. The new family of Neugents is part of a revised CA product named "Jasmine ii," an e-commerce software package that allows businesses to link customers, partners and suppliers.

Neugents work by collecting data and analyzing it every way imaginable to find trends, said Dave Johnson, CA's vice president of information management. The technology learns the normal behavior of a computer network, for example, and can figure out when problems could occur, he said. "It's software that thinks the same way a human brain learns and uses information."

IT executives at Brigham Young University says the technology works as advertised.

During the past year, the college had to increase its pool of servers but didn't have the money to increase its computer staff, said Sorrel Jankins, the school's director of server engineering.

Instead, the university installed Neugents to help monitor the network.

"If you give a technician something to look at it, he will pretty much look at some key areas and hopefully trip over the problem. Neugents does the same thing but looks at tens of thousands of scenarios," Jankins said. "It's given us very accurate predictions of network outages."

Charles Tyrwhitt Shirts, a London-based clothing company, is using Neugents to help develop its marketing strategy.

In the past, the company had shipped marketing brochures to its 250,000 customers about eight times a year, hoping to boost sales. But with Neugents, it can target its advertisements to customers who are most likely to buy products, said James Stewart, Tyrwhitt's vice president of U.S. marketing.

"Neugents enable us to market to customers when they're likely to buy again by looking at their previous purchase patterns," he said. "Some of our customers buy from us during a sale, some just buy white shirts, and some more affluent ones buy our higher-end shirts. We want to market to people to the way they spent money in the past."

The clothing company expects Neugents to help it save money in marketing costs while increasing sales.

While testing the technology, Tyrwhitt recently input all customer information in 1998 and asked the Neugents to predict which 10,000 customers would be good customers in 1999. After comparing Neugents' results with the actual 1999 sales results, Stewart found the Neugent technology had a 55 percent accuracy rating.

Although 55 percent isn't stellar accuracy, Stewart said the results were good enough.

"It seems like a low percentage, but we would normally send that offer out to 250,000 people and only get 2 or 3 percent (response)," he said. "By narrowing it down to 10,000 people and getting that accuracy, (it) means we saved a fortune on marketing."

Illuminata's Eunice said the technology will improve and become better at predicting the future as it receives additional data.

"If you can guess somewhere between 70 to 90 percent of what a likely purchaser is, that's pretty impressive," he said.