Questions about enterprise artificial intelligence for banks are coming as news of fraud in stimulus programs spreads. Banks that protected themselves will appear far-sighted. That’s how more, not less, transparency about fraud detection and prevention efforts just might wind up leading to greater profit now and in the long-run.
An enterprise data and AI capability can demonstrate to regulators, investors and customers that the bank knows what’s going on within its servers and networks. Done right, machine learning solutions can hyperscale and improve with experience. The more data they ingest, the smarter they become.
So how about stimulus fraud?
Auditing and transparency
When it’s implemented correctly, enterprise anti-fraud AI should analyze all newly arriving data, identify changing patterns, and suggests updates to segments and rankings based on new information. As a result, it readily identifies subtle patterns suggesting emergent behavior for consideration by subject matter experts. Further, the more data sources available, the better the grouping that results from fraud-detecting behavioral segmentation.
More importantly, good anti-fraud AI technology does not require labeled data to derive an initial segmentation. Removing the requirement for labeled data permits substantial expansion of the number of data sources, including customers of a bank’s customers (KYCC).
By pursuing this kind of rigor, anti-fraud AI should provide complete transparency into what drives the segmentation. Enterprise quality AI should produce a complete documentation workflow containing simple decision trees that can be shared with internal model governance boards and with external regulators. Decision trees are excellent ways to visualize complexity for regulators and internal model review boards and are a key part of the justification step in anti-fraud.
With this in place the bank can better communicate and demonstrate to regulators, customers, investors and policymakers how it is distributing funds and catching wrongdoers. This is particularly helpful when news organizations start receiving lists of stimulus funds recipients – sometimes lists with critical flaws – and start hounding banks for answers.
Daily checking
To keep up with fast-moving events, high quality anti-fraud AI should analyze customer transactions daily. It should automatically generate lists of, and can alert against, customers showing changes in behavior over time, such as the customer’s behavior deviation over time; from their norms, their behavioral peers, their past and their industry. The changes in a party’s behavior compared to their peers in their segment is important. The deviation in customer behavior compared to the information provided during KYC is also key. Deviation from nature and purpose elements should be monitored. Party migration between and across segments should also be tracked.
Knowing which behaviors, scenarios and typologies your system’s rules currently address is only part of the management challenge. Every day, changes to products, geographies, regulations, acquisitions and source data can undermine the work you performed in your prior tuning exercise. This leaves you exposed to risks from those new and emerging behaviors.
Enterprise AI anti-fraud should provide detailed, auditable reports to highlight emerging behaviors and further, the existing rule applicability to immediately address them, providing detailed segment characteristics and membership insight. Behavioral segmentation provides insights to investigators about changing party behaviors.
A steering wheel
An intuitive and insightful human user interface is needed. It should be driven by an easily integrated alerting engine, mark out any risk, be capable of being digitized, and can be discovered, alerted, and sent to case management. It should be visualized, investigated, escalated, added to a watch cycle, automatically create a segment for subsequent monitoring, submit data to any auto CMS/SAR/STR system.
The bank should be able to discover not just fraud but precursors like cyber attacks and attempts and the inevitable money laundering that follows. It should be able to discover and alert on everything from tax evasion to trafficking. New enterprise risks should be identified at the party and entity level and be auto alerted and visualized, contextually, for confidence and peace of mind that an institution is fully empowered and prepared.
Ensuring that you are fully covered for all known and unknown, knowable and currently unknowable risks. New entity risk detection, provides a summary of all risks in a single view, enabling instant visualization and machine or human prioritization, in line with your institution’s appetite for risk and backs it up with deep, drillable, pre-fetched, pre-aggregated and enriched party data. Account behaviors, credits, debits, payment histories, payment flow visualizations and more are all available to give a holistic and clear picture to your investigator and analyst community.
But what’s all this transparency amount to, apart from being a feel-good idea?
Transparency is more than nice – It’s the foundation of trust!
Harvard Business School’s Ryan W. Buell details the benefits of operational transparency. In a nutshell, if you have the capacity to offer a window to all stakeholders, into how services are delivered, it can dramatically boost the perceived value of those services.
Examples across industries are straightforward and convincing. A diner who can see and talk to a chef values the food more. A person searching online for a flight is more loyal to a site that indicates the number and names of airlines it is checking. Customers are more patient with an ATM machine that reveals the steps it undertakes — contacting the host bank, accessing the account, counting the money — than with one that merely states, “processing.” The concept works in reverse too: when employees have contact with customers, they learn from the interaction and are motivated by the enjoyment of making a difference in people’s lives.
If you have right enterprise AI solution, you’ll have trust. That’s the stuff great brands are made of.