This paper describes the Natural Language Processing (NLP) component of an e-mail monitoring product called Assentor. Assentor monitors electronic correspondence for brokerage firms. It uses pattern-matching-based information extraction technology to find and quarantine e-mail messages that indicate, among others, customer complaints, insider trading, stock hyping, hard-pressure sales tactics, and firm preservation issues such as jokes and obscenities. This paper presents a quantitative evaluation of applying pattern matching vs. keyword-based searching to e-mail monitoring. Our evaluation shows that pattern matching performs significantly better than keyword-based searching both in terms of recall (false negatives) and precision (false positives).