Financial services firms are required to monitor client activity, identify potential financial crime, and investigate further where necessary. They typically use detection scenarios to generate monitoring exceptions for follow-up by risk owners and financial crime compliance within the firm. But IT programs that use logic and rules to monitor transactions can only achieve so much. The problem with detection scenarios has been that firms use the same scenarios to monitor the transactions of vastly different customers in widely different countries. It makes sense for a detection scenario to flag cash transactions over the account of a financial institution as potentially unusual, but it does not make sense for an account that belongs to a retail business over which cash transactions are the norm. If detection scenarios are not customized by country and by business line, a firm's transaction monitoring will be burdened with false positives, wasting costly finite risk management resources.

Ideally, detection scenarios should be unique to each customer account, and machine learning holds the promise of achieving this in the future. In the meantime, financial service providers are starting to employ detection scenarios that are customized by business line and by country. There is a problem here though if the detection scenarios are known to all employees at the firm. Criminal gangs could use insiders in the financial services industry to help find pathways for circumventing control policies and procedures. As a result, it is typical for detection scenarios to not be made available to all employees through a company's intranet, but rather they are restricted and only made available to certain employees on a need to know basis.