Fintech firms aiming to sell generative AI services to banks would be wise to heed the AI fears, concerns, and preferences of bank executives and their corporate board members.

The reason:  Ninety six percent of senior bank leaders say they are far more involved in technology and IT purchase decisions due to increased interest in generative AI.

Among the top concerns likely to impact fintech purchases at banks: Data security issues such as the leaking of confidential corporate data via generative AI large language models and AI regulatory compliance challenges related to the workings of such models.

These are among the key findings from a new survey on the outlook for generative AI adoption, prepared by the Harris Poll organization on behalf of Google Cloud, providers of cloud services to facilitate the deployment of AI models.

The survey – conducted online in October – sought to explore the sentiment towards generative AI among North American banking executives and consumers and in doing so, provides guidance for fintech firms seeking to serve up Gen AI models and capabilities to banks.  

It is based on a survey of 350 banking executives responsible for AI decisioning and more than 2,000 banking consumers in the United States.

The good news is that a significant number – 47% of banking executives – say their banks have taken the plunge and are in the proof-of-concept stage of generative AI implementation while 35% say they are currently piloting and testing use cases.

“Banking leaders are no longer just experimenting with gen AI; they are building and rolling out use-cases that can improve operational efficiency,” said Yolande Piazza, VP of Financial Services, Google Cloud in a press release.  

At the same time, a host of fintech firms have sprung up to offer AI technology specifically designed to assist banks and related financial activities, among them: Arteria AI, Boosted.ai, Bud Financial, Hazy, and Unit21.

However, it remains to be seen how many pilots or test cases involving Gen AI withstand the test of time. 

“There’s a big difference between conducting AI experiments and actually moving into production,” says Christine Livingston, managing director of artificial intelligence practices at consulting firm Protiviti.  She notes that she is starting to see elements of “AI fatigue” due in part to the extended periods of risk management and governance reassessment required for such projects within the banking sector, a highly regulated industry.

The survey reports that current implementation at banks includes the use of generative AI to summarize complex financial information (49%);  to summarize capital market research for client briefings and to facilitate faster investment decision-making (49%); to enhance chatbots and virtual assistants for customer interactions (48%) and for predictive modeling of risk scenarios (40%);

Other findings are that banking executives believe the top benefit that generative AI efforts can bring to the industry is operational efficiency along with cost savings (49%), with the expectation that it will provide better data analysis and predictive analysis (45%) and improved fraud detection and security (44%), all critical activities for financial firms.

Generative AI is also expected to drive revenue growth by improving investment research (41%); providing more effective marketing or customer segmentation (38%) and better customer acquisition and retention strategies (38%).

However, the challenge for many fintech firms hoping to overcome any reluctance on the part of bank executives to adopt generative AI capabilities will be to address any perceived drawbacks or challenges. 

According to the survey, these include:  Concerns on the part of national banking executives around data security issues and specifically the possible leakage of confidential and valuable company data to the AI language model (56%);  concerns about compliance uncertainty in the days ahead in an industry that is highly regulated (39%);  as well as the need to invest in appropriate technology to get bank data into the cloud (49%) and the need to update AI data policies (49%);

Consultant Livingston sees other challenges as well:  When it comes to gen AI capabilities, “I see a lot of hesitancy around its use for customer-facing applications;  You need very specific guard rails in place so that AI models do not provide inaccurate information to customers,” she says.

 She adds: “Having purpose-built generative AI capabilities are the most likely to be successful,” when approaching banks and these offerings “need to complement and extend the bank’s core architecture stack,” for example, augmenting or extending the fraud detection capabilities already in place.

Finally, “Fintech firms really need to be transparent and accurate about their generative AI capabilities,” as opposed to just using the tagline.  “Be prepared to explain your models and data,” so that you can highlight their applicability to specific business needs and assist banks with their compliance and model transparency requirements, Livingston says.

Not surprisingly, executives at Google Cloud are upbeat about any adoption challenges.  According to Zac Maufe, global head of Regulated Industries at Google Cloud:  “This recent research reinforces what we’ve been seeing in the banking industry for the past six months, which is that Gen AI can represent massive productivity and operational efficiency opportunity,” and one that banks will surely want to pursue on behalf of their customers.