Chuck Price

Organizations are increasingly seeking to optimize and enhance customer experiences by leveraging AI to improve personalized marketing. The banking sector is no exception, with most financial institutions at least exploring the potential if not already deploying AI. 

Benefits of AI in Optimized Offers

Optimizing offers for banking products benefits both customers and financial institutions. By leveraging AI, banks can analyze vast amounts of customer data to tailor offers that align with individual needs and preferences. This level of personalization has the power to enhance customer satisfaction and drive engagement with relevant products and services. Furthermore, AI can streamline and automate processes, leading to operational efficiency and cost savings for organizations.

Risks and Privacy Concerns

While the benefits of AI are enticing, it is crucial to address the ethical risks and privacy concerns associated with its implementation. AI relies on large volumes of training data, which is the source of an initial set of considerations.  Some risks include:

1. Security.  Customers entrust financial institutions with their sensitive information, and it is paramount that information is safeguarded with the highest levels of security. Unauthorized access, data breaches, or improper use of personal information could have severe consequences and erode customer trust.  
2. Minimization.  Machine-based approaches consume large volumes of data to gain deep insights and unlock nuances that were invisible to traditional analytic models.  Without careful planning, this can increase the risk to people by allowing processes and systems access to more data about each person.  This may create multiple vectors for fraud in systems less hardened against attack than operational platforms.  One simple tactic for limiting these risks is to minimize the amount of data to only what is most critical to automated decisioning – and eliminate all ancillary data from AI processes.
3. Anonymization.  The final and perhaps most important consideration is to use strategies that disconnect relevant data from the identity of the person involved.  This should already be a standard practice in banking, but as AI pulls in additional sources, it is vital to carefully inventory and review the list of data elements made available to AI.  Techniques from simple one-way hashing of identifiers and stripping of publicly identifiable information in analytic feeds to leveraging modern data clean rooms can provide an effective layer of protection by ensuring the data used to train AI cannot be traced back to any individual person.

Prior to deploying AI, it is important that the institution revisit its corporate data security and governance policies as well as analytic environment controls.  It is very likely these will require updates to account for increased risk.

Transparency 

Data is not simply a business asset.  When that data is linked to an identifiable person, it is an extension of that individual.  With AI comes the ability not just to process greater quantities of data but also the ability to embed automated decision-making into more processes.  This has already caught the attention of the FTC, which has published an advanced notice of proposed rulemaking to solicit public comment on commercial surveillance and data security practices that harm people – which was shortly followed by the White House release of an AI Bill of Rights blueprint.  

Although eventual federal regulations are likely, organizations should not wait to act responsibly.  They should consider clearly disclosing:

1. That automated decisioning is being used
2. The types of decisions (including offer targeting) that are made
3. The types of data made available to automated decisioning
4. The benefits to the person

These disclosures should be accompanied by a point of contact within the organization and any relevant reminders of consumer data protection rights.  Companies’ legal teams should be involved to ensure their disclosures are ready for an AI deployment. Similarly, organizations need to be prepared to explain why a particular decision or outcome occurred or unexpectedly, did not occur.

Impact on Human Decision-Making

While AI can greatly enhance offer targeting, it is crucial to strike a balance between automation and human involvement. The overreliance on AI without appropriate human oversight can lead to accountability issues and potential harm to people. It can also lead to an erosion in the understanding of how business decisions are being made.

Organizations must specifically design AI implementations to address this by incorporating human oversight and intervention that will maintain ethical decision-making standards, prevent unjust outcomes, and continuously monitor the flow of data through AI processes.  Without this, control of personalized marketing decisions will be lost to AI, which may begin optimizing based on metrics or correlations that are less understood over time.

Conclusion

As organizations embrace AI to enhance offers in the financial services sector, it is paramount to prioritize ethical considerations. While AI offers enticing benefits such as personalized experiences, lower acquisition costs, and operational efficiency, organizations must address the associated risks and responsibilities. Identifying ethical considerations begins not with AI itself but with the technical and regulatory environment in which it will be deployed.  It is critical that privacy concerns, transparency, and the impact on human decision-making must be focal points before deploying AI. By prioritizing data security, striving for transparency and explainability, and maintaining human oversight, organizations can harness the power of AI while upholding ethical standards. The banking industry can lead the way in utilizing AI to responsibly optimize marketing offers.