An expert system for realty loans; Dominion Bankshares testing PC-based software.

An Expert System for Realty Loan

Dominion Bankshares Corp. is testing new software aimed at giving personal computers enough smarts to determine which real estate loans are most likely to go bad.

The Roanoke, Va., banking company plans to use an artificial intelligence system to analyze real estate and construction loans, the types of credits that have ravaged other banks' balance sheets.

Dominion expects the software to act like a canary in the coal mine - an early warning system that alerts lenders to potential credit trouble under changing market conditions.

"This doesn't mean we won't have loan losses," said Charles Abbitt, senior vice president. "But the system is going to give the account officer sufficient warning that there is trouble down the road, so he can address the situation."

New View of Real Estate Loans

Dominion is the first bank to apply this type of advanced software to the problem of real estate and construction loans. The bank plans to have its 60 real estate and construction lenders using the personal computer-based system by early next year.

The $10 billion-asset bank joins the small but growing list of financial institutions that are experimenting with artificial intelligence software. Perhaps two dozen banks have using been AI in commercial lending. Virtually all of these early adopters began to explore the software when they realized how fast the non-performing loans were mounting.

Dominion's software is known as an "expert system," which is designed to contain the knowledge of seasoned professionals in particular fields.

In the opinion of some bankers, expert systems can help them avoid the pitfalls of bad decisions.

Choice of Forecast or Analysis

Banks have taken two approaches to applying artificial intelligence software, either to forecast the likelihood of specific loans going bad or to analyze the overall financial stability of corporate customers.

For example, Chase Manhattan Corp. is using another type of artificial intelligence software called neural networking to forecast which loans will go bad in three years. Neural nets, as they are sometimes called, are designed to spot patterns without being told what to search for.

A handful of banks, including units of Security Pacific Corp. and First Interstate Bancorp, bought a comprehensive, mainframe-based expert system for credit and financial analysis from Syntelligence Inc. Syntelligence went bankrupt last year. The software is now owned by a consortium of companies.

Other banks use commercially available AI software for financial analysis. Swiss Bank Corp. developed its own system for credit evaluation, which it plans to install in New York.

Looks at Risk

Dominion is working with Data Select Systems Inc., a software company in Woodland Hills, Calif., to develop its expert system. The software has performed well in tests and has been used to evaluate some loans for apartment buildings, Mr. Abbitt said.

"It helps you look at the downside risk," Mr. Abbitt said. "There are so many things to look at in a real estate loan. This pulls it all together."

Lenders at Dominion currently perform the loan analysis using calculators and spreadsheet programs running on PCs. But to try to determine if a loan needs to be redone in response to, say, disappointing vacancy rates, may take two days. The new system may cut that evaluation time to hours.

Dominion has about $1.3 billion in real estate and commercial loans. While its $13 million in nonperforming construction loans is small compared with its peers, the bank wanted to take action against possible sizable losses in the future.

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