In the critical proof of deposit (POD) stage the face value of checks is compared with the deposit statements and entered into the bank's computer database. The bank encodes or prints the correct dollar amounts on the bottom of the checks, bundles them into groups (cash letters), and delivers them to the other banks via the Federal Reserve system.

A data entry clerk scans and enters the check dollar amount (courtesy amount). The deposit slip is scanned and matched against the adding machine tapes of the deposits the teller received during the shift. If the deposit slip and courtesy amount don't match, the check is kicked back for a closer look. Or, in some situations, minor variations may be ignored, or the check amount may be defaulted to one number when the amounts entered are close.

In the conventional POD process, an encoding machine sends the checks to employees who enter the dollar amounts. This system is simple, relatively failproof and provides a good means of checking hard to read numbers. When the clerk can't decipher the courtesy amounts, they can simply read the handwritten, scripted dollar amounts. The cost of processing checks can be enormous, especially for bigger banks that process millions of checks every month. Labor costs are high and, because of the varying workloads, the work is very hard on data entry personnel. Federal law dictates that checks without sufficient funds to cover them must be returned to the issuing bank within 24-hours or the bank of deposit must cover the loss. Now, just imagine a situation where there is a heavy snowfall at the end of the week before a four day weekend.

The following Monday there could easily be four times the normal workload. Usually data entry personnel can't leave work until all the checks are processed (otherwise, the banks would have to swallow the cost of all the bad checks it receives). Not surprisingly, the turnover rate in this line of work is quite high.

The concept of image-enabled POD is relatively simple: give the computer first crack at recognizing the courtesy amount and deposit slip amounts of the check. After all, we live in the age of microprocessors where computer time is cheaper than human time, isn't it?

The software that recognizes the handwritten numbers on checks is a special- purpose intelligent character recognition engine called a Courtesy Amount Reader (CAR). Checks and deposit slips are sent through a reader/sorter that prepares check images and deposit slips for the CAR engines, which are engineered to read the courtesy amounts and extract the correct numbers from the chicken scratch entered there, no easy task. In fact, the very best CAR engines on the market boast 40 to 60 percent success rates.

This is simply not accurate enough for many document types. Would a legal secretary, for example, bother to correct a document recognized with only 80 percent efficiency? Not likely - it'd be much faster to type the document from scratch!