As a leading digital agency specializing in the financial services space, ZAG recently interviewed 24 employees from 14 different banks and credit unions to understand not only how they’re implementing AI, but how they are doing it responsibly. These institutions had varying AI maturity levels - from no AI plan to full AI governance teams driving multi-year strategies.
We spoke to different employee levels and departments spanning junior to Executive, across Marketing, Digital, IT, Development, Communications, and Brand. While there were many different viewpoints and approaches discussed, there were also many synergies. It was clear that AI is viewed primarily as an efficiency tool, but it requires human oversight, trust isn’t freely given, and privacy concerns are always top of mind. This article dives into how institutions are balancing these efficiency gains, keeping information secure, staying authentic and ensuring accuracy.
How are Banks and Credit Unions approaching AI today?
The answer is slowly. Banks and Credit Unions know better than anyone the importance of privacy and protecting personal information at all costs. Here are some key findings regarding how institutions are using AI while keeping data protected.
- Banks and Credit Unions are being methodical in their approach. The institutions interviewed with higher AI use and comfort have not implemented it blindly. Rather, they have identified the strategic areas where AI is most beneficial, then they built out processes in controlled environments ensuring to always protect sensitive information. These adopters are measuring the success of using AI and are constantly looking forward to other ways AI could be used to support their goals. While many AI tool variations were discussed, embedded AI was favored. At a minimum, all institutions agreed that AI tools must operate in a closed environment, to reduce the risk of data leaks and prevent data from being used to train other AI models.
- The Institutions who did not have AI strategies in place ran a higher risk of unintentional misuse and fragmented adoption. These banks and credit unions fundamentally understood what types of information should and should not be inputted into AI, but the lack of an official strategy created a disjointed approach which limits effectiveness and creates risk.
- AI adoption varied across departments. Some departments had access to specific AI tools while others could not use AI in any capacity. In a few instances, AI access was limited to just a handful of people on a team resulting in inconsistent usage. For example, Marketing teams were more open to using AI than IT / Security teams were, even though the concerns and perceived risks were similar.
So, how is AI actually being used by Credit Unions today?
Efficiency emerges as the top benefit
There were a lot of different areas where AI is being used or being considered for use. The top applications mentioned included:
- Analyzing and pulling insights from large datasets
- Enhancing customer/member experience
- Detecting fraud
- Writing code
- Idea generation
- Creating first drafts of content
- Building internal and external knowledge centers
- Recording videos or audio
One common thread was that efficiency is where everyone agreed is AI’s top strength.
AI is significantly helping financial institution teams move quicker, generate more output, and reduce managerial work. It doesn’t mean AI is replacing human employees, but it’s allowing teams to focus on more strategic priorities. In short, AI allows teams to produce at a quicker pace, saving institutions resources and money.
Where AI adoption isn’t so simple
While AI drives significant value, it doesn’t come without risk.
- Data protection: Data protection is the number one area of concern from all institutions interviewed. The personal and financial information Credit Unions and banks have access to is significant, and the fear of this being leaked or mishandled by AI is the top reason non-AI adopters choose not to implement. These privacy concerns also extend to proprietary or sensitive company information. Because of these risks, open AI models for the most part are a “no-go” and AI adopters are choosing closed models that help mitigate these risks.
- Inauthenticity: Marketing and Brand teams were primarily concerned with appearing inauthentic, especially as AI has been known to completely falsify information or in some cases creating images where people have too many fingers. In these cases, AI is only used at the beginning of the creative process. If used at the end then there are special human reviews conducted to ensure the output is correct. For institutions who pride themselves on being relationship driven and focusing on human interactions, there is a fear that releasing materials which are clearly generated by AI can be harmful to their carefully cultivated brand.
- Ownership: If AI generates content, then who owns that content? Could institutions be forced to remove this content in the future and will copywriting issues arise? What is stopping AI from using the same generated content with a competitor? These risks are real, there is not a clear answer yet.
So how are institutions mitigating these risks? From a data privacy perspective, closed models are preferred. Institutions already using AI have policies in place and are teaching employees how to use AI and what information can be entered into AI vs what cannot. From a Marketing standpoint, teams are generally using AI as an idea generator, with human oversight before considering the end product finished. These approaches effectively allow teams to use AI to drive efficiencies while retaining that authentic human touch. This also ensures AI accuracy, while helping to preserve ownership of the output.
What ZAG's Research Ultimately Shows
If there is one thing this research makes clear, it’s that responsible AI use means AI cannot be used without significant oversight. Banks and credit unions should incorporate AI into their processes with planned intent and supervision. Deciding where AI can drive efficiencies and putting policies in place ensures employees understand how to use AI effectively while protecting data at all costs. Lastly, and the most important, is that human involvement is critical. Institutions who had the most success with AI did so because it was supporting - not replacing humans. AI can do a lot of remarkable things but still requires judgement and oversight to ensure the final output is aligned with what the institution stands for.
