AI in AML: From Experimentation to Operationalization

Create a vendor selection project & run comparison reports
Click to express your interest in this report
Indication of coverage against your requirements
A subscription is required to activate this feature. Contact us for more info.
Celent have reviewed this profile and believe it to be accurate.
8 May 2019
Arin Ray

Early last year we predicted 2018 to be a critical year for operationalization and wider adoption of artificial intelligence (AI) and machine learning powered solutions in Know Your Customer (KYC) and Anti-Money Laundering (AML) operations at financial institutions. The developments of the past year proved that right, and the pace of adoption even surpassed our expectations as we see banks of different sizes, types, and from different jurisdictions are moving on from pilots and proofs of concept to using AI in production.

These developments are a natural outcome of the growing costs, inefficiencies, and sub-optimal outcomes plaguing AML operations at most banks. Regulators and supervisory agencies have started to take note and share generally positive views and feedback. A recent joint statement from supervisory agencies in the US is being regarded to be a catalyst for further promoting innovation and expediting adoption of AI and ML solutions in AML.

There is great appetite among industry participants to learn about live use cases of AI being applied in AML, as well as the challenges and considerations when operationalizing AI. In a recent Celent report we analyzed in detail live AI solutions from eleven vendors representing a good mix of incumbent providers and industry newcomers, as well as a mix of solutions that span different parts in the KYC-AML value chain. The vendor list includes: Arachnys Information Services, Ayasdi, BAE Systems, Fiserv, IBM, NICE Actimize, Pelican, Quantaverse, Quantexa, SAS Institute, and ThetaRay.

It is interesting to note that the benefits of applying AI and machine learning in AML can come in several ways:

  • Reducing false positives is a huge area for operational improvement and is also the most commonly sought after. Vendors and banks have reported use of AI and machine learning are helping them lower their false positive rates by over 30%; some have reported much higher estimates.
  • False negative identification is another critical improvement enabled by AI and ML. Some banks have reported identification of previously missed cases as well as identification of previously unknown relationships and behavior.
  • Productivity improvement is another benefit and can come in several forms. Banks are able to speed up investigation time, typically by 50% or higher, by automating much of the manual search and data entry process, eliminating redundant tasks or reworks, quickly identifying duplicate alerts, automating tasks for data gathering, negative news lookup, and so on. Another metric for productivity improvement comes in the form of better escalation rates, where upwards of 30% improvement has been reported.
  • There is also significant potential for reducing AML program costs. Efficiency and productivity improvements reduce alert handling costs, while AI and ML tools can eliminate the need for periodic tuning exercises, thereby reducing one-off but significant costs and efforts. We have come across estimates of over 30% and several million dollars in potential cost savings.

We will continue to track the developments and further adoption of AI solutions in AML and publish research as adoption grows and new use cases emerge.

Insight details

Content Type
Blogs
Focus
Risk Management & Compliance
Location
Asia-Pacific, EMEA, LATAM, North America