Home / Banking Strategies / Dirty data is not an AI-for-AML dealbreaker

Dirty data is not an AI-for-AML dealbreaker

New technologies can contend with many of the data-quality issues that would have tripped up older compliance systems.

Jan 14, 2022 / Technology

A rise in financial crime since the COVID-19 pandemic and the burden of increasingly complex compliance requirements mean that many older anti money-laundering (AML) systems are not coping well.

Artificial intelligence is emerging as a potential game-changer. With its ability to rapidly process and spot anomalies in large volumes of data, AI could provide AML technologies with an upgrade and could unlock billions of dollars  in value globally.

Many believe that, in order to reap the full benefit of an AI transition, organizations need high-quality data. But using dirty data as an excuse to shy away from AI to combat financial crime could prove riskier in the long run. Regardless of data quality, there may be potential benefits from an AI transformation.

How clean is clean?

What you believe to be dirty data might not be as dirty as you think. Anomalies like duplicate entries, misspelled or alternatively spelled words and names, punctuation errors, and some types of incomplete or outdated data that trip up older systems tend to not be a problem for AI. Advanced AI solutions are programmed to analyze data from multiple angles and sources, so they can produce meaningful results even in the presence of anomalies and exceptions.

For example, a person making a transfer could write their full first name one time and only their initials the next time. While less advanced systems would flag this as an anomaly, AI can corroborate the two entries and “understand” that they apply to the same person. The same goes for numerous other data discrepancies that could be keeping you away from AI but are not as problematic as you first thought.

Start small

The average bank processes thousands of transactions from multiple sources each second, so the scope for data error is staggering. It’s no wonder that organizations worry about dirty data. An AI transformation can be done in phases that allow lessons learned in one area to be applied in other areas in the future.

Isolating SWIFT data from the masses of other data can be an ideal place to start. An advanced AI-based system can rapidly analyze terabytes of data to reveal suspicious transaction profiles and unlock hidden insights. Implementing AI in this one area alone could save significant compliance costs, reduce human error and free up hours of personnel time.


The changing intersection of banking and technology

Banks and fintechs are on the partnership track

U.S. banks are playing catch-up on digital technology

Moving from low-code hype to successful implementation

Analyzing banking data and putting it to good use

Reap the rewards

If your organization is truly plagued by dirty data, then any rules-based AML system will be compromised. Just “living” with dirty data is, therefore, a dangerous option, so cleaning up your data as much as possible should be a priority. Adopting an unbiased or intuitive AI system may offer value from the get-go even as your data cleanup continues in the background.

  • Fewer false positives:  Legacy AML systems are notorious for returning false positives, which increases operational costs. Banks may be tempted to make the rules less sensitive, but this would expose them to greater potential risks. A rule-agnostic AI system cuts this risk by relying on data or behavior rather than predetermined assumptions.
  • More patterns identified: Intuitive or unbiased AI discovers patterns undetectable by traditional systems because it effectively looks at data with “fresh eyes” every time.
  • Virtually unhackable: Rule-based AML systems based on human judgment, rather than absolute data, are more vulnerable to attack. It is relatively easy for determined criminals to uncover the rules and find ways to bypass them. Rule-agnostic systems do not have the same exposure.

New AI technologies can bypass many of the dirty data issues that would trip up the legacy systems. This means that you don’t need to wait until you’ve done a thorough data-cleansing to start protecting yourself against would-be money launderers and other criminals.

Idan Keret is chief customer officer for ThetaRay.