In banking and technology, ChatGPT and other types of generative AI is on most people’s mind. In talking with clients they are starting to explore how generative AI can be leveraged to enhance their financial crimes prevention processes. Given that executives
are under pressure to reduce costs while managing risk a very logical question to financial crimes executives would be how can generative AI be used to be more effective and efficient.
If I were still at a financial institution I would pose the answer a different way. Yes ChatGPT looks very promising especially considering that it was launched in November 2022. First I would assess the current use of automation and AI within my financial
crime case management process.
Alert and Investigator Views
How is the relevant data of alerts and investigations being gathered for analysts and investigators to make risk decisions. Are there currently manual processes that are creating a swivel chair experience for the users? Analysts and Investigators skills
are to make risk decisions not gathering information. This is a clear baseline use for automation.
How is Work Assignment Being Managed
Is the work being manually assigned by a manager or are users pulling from a work queue? A quick lift would be establishing some business rules for smart routing that allow users to work on items that are best for their experience, geo, etc. Regardless
there should be a unified intake and management for the work items that can track is working on what, the time spend for each action and what is past it’s Service Level Agreement. Having visibility into the bottlenecks will allow for further efficiency and
automation.
For example, a large bank in Asia Pacific automated a number of manual processes by automating certain aspects of their financial crime case management which allowed them to reduce their overall investigation time by 75%, manage alerts more efficiently
and holistically.
Work Assigned by Predictive AI
Once you have maximized your lift from smart routing the next lift is from work assigned by predictive AI. This involves moving from a rules based assessment to AI that can assess the work and predict the risks, complexity, and patterns and route the items
to the user who has the best skillset for these items. This AI should be transparent and explainable to both internal and external parties.
Actions Assigned to Work Items
Not all alerts/work items present the same risk to financial institutions. Financial crime professionals need to way the risks along with operational efficiencies. We have seen some financial institutions starting to embed guided processes into their financial
crime case management workflows. For example, they may factor in the risk rating of the alerts(s), line of business, etc and assign a number of tasks associated with the risks of these items. So a lower risk set of alerts will have a subset of steps that
are embedded in the workflow that the investigator needs to take. Whereas if it is a higher risk then there are additional steps added by the workflow. Of course if an investigator sees something worth investigating further then they can do additional steps.
By having guided processes it ensures that investigative steps are being uniformly followed for the risk presented while balancing operational efficiency.
Should financial crime prevention areas be exploring how generative AI can be used to enhance their processes…absolutely. However it is still early in the journey with generative AI. Financial crime prevention areas have spent the past three to five years
on fine tuning their detection systems to reduce false positives and manage risk more effectively. As the global economy continues in this uncertain economy period, financial crimes professionals are going to be under a lot of pressure to continue to manage
risk while do it with the same and less resources. Giving a deep look into what areas of the process are still manual or lightly automated can provide the opportunity to leverage automation and AI to make the financial crimes case management process more
efficient and effective.
So yes explore ChatGPT and other types of generative AI. However look to technologies that are proven and continuously enhanced and make a business case to the executive(s) who are asking these questions. By using already proven technology allows for the
reduction of risk and increasing the overall success rate.
