Comerica Beefs Up Trust System

Using artificial intelligence software and an easy-to-read data base of customers' holdings, Comerica Inc. has developed one of the most advanced investment analysis capabilities of any bank.

Comerica says the automation, which cost about $1 million, has already improved its ability to manage money.

For the past two years, Comerica Capital Management Inc., a unit of the Detroit banking company, has been building computer systems and installing powerful workstations on the desks of money managers. The project is more than halfway completed.

Money Managers Run the Show

What's more, Comerica's money managers have been running the project - one of the more complex in banking - rather than its technologists.

James A. Sansoterra elected himself to run the project. Mr. Sansoterra is a vice president with Comerica Capital Management, a captive investment company that manages about $12 billion in assets for the trust customers of Comerica banks.

"I was a money manager, but I was not happy with the way I was managing money," he said. He focused the project on improving the performance of money managers.

The combination of computer systems and personal work-stations has enabled Comerica's money managers to customize the portfolios of trust clients.

That approach is a step up from the typical investment approach of classifying customers together under a few standard categories that reflect investment goals.

The customer data base and computerized stock-picking system are finished. About 60 money managers now tap away at workstation keyboards, perusing account information and making trades.

By yearend, 100 money managers will have workstations linked through computer networks.

Improvement Noted

Mr. Sansoterra said he already sees an improvement in his managers' performance. "We have a more consistent approach. More accounts with similar investment objectives are getting similar returns."

Partly because he had no stake in which brand of computer was used, Mr. Sansoterra made some unusual technology decisions. He chose Sun workstations, which run on a Unix operating system, a software used by a wide variety of computer makers.

Previously, virtually all systems in banks ran on special software unique to International Business Machines Corp.

He also built a large-scale relational data base, something few banks have. In a relational data base, pieces of information are stored separately, not linked, so scouting for a common piece of data on 10,000 customers is easier than in more standard designs of data bases.

Comerica Capital Management maintains a data base on 10,000 trust customers - their assets held, state of residency, tolerance for debt, related accounts, and other information.

The bank built its customer file using a data base management software from Sybase Inc., Berkeley, Calif.

The customer data base encourages investment managers to examine individual accounts. Say a money manager discovers IBM's stock is down 6%. From his workstation, he can search the data base to find out which customers have the cash to buy the stock. The list of customers comes up within seconds, according to Mr. Sansoterra.

"No bank has a data base you can cut into like this," he said. "Banks are in the 12th century when it comes to this kind of technology."

A money manager on the phone with a client can get the current value of his holdings - based on the previous night's stock prices - within 10 seconds. Comerica is improving that to reflect the minute-by-minute value of the portfolio.

Customers can get a copy of their holdings faxed to them immediately.

|Does the Work of Six People'

Every day, the customer and portfolio data are run through a homegrown artificial intelligence program, which reviews a client's portfolio in light of investment objectives. If the holdings do not meet the objectives, the system will recommend trades. Those recommendations are almost always taken by the money managers.

"By itself, the artificial intelligence system does the work of six people," said Mr. Sansoterra. In the year and a half the system has been crunching through portfolios, it has done 110,000 trades without a glitch, said Mr. Sansoterra. The savings from this software alone will pay for the rest of the automation, he said.

Other banks do computerized stock picking. But no other trust bank combines computerized stock picking and allows trade decisions to be made as quickly.

Subscribe Now

Access to authoritative analysis and perspective and our data-driven report series.

14-Day Free Trial

No credit card required. Complete access to articles, breaking news and industry data.