25 questions financial institutions are asking about gen AI
63% of financial services respondents have generative AI in production
Financial institutions are exploring generative AI through their own center of excellence initiatives, pilot programs and internal applications.
However, surrounding the technology “there are more questions than there are answers,” Kris Lazzaretti, senior vice president of data analytics at payments data solution provider Deluxe, told Bank Automation News, noting that the industry is in the early stages of experimentation but application “is promising.”
Google Cloud and National Research Group surveyed 340 senior leaders of global enterprises in the financial services industry about generative AI experimentation and applications. The September study found that:
- 63% of financial services respondents have generative AI in production;
- 35% of respondents continue to evaluate or test generative AI; and
- 26% of respondents have been using generative AI in production for over one year.
Generative AI is deployed in the following areas at these institutions:
- Customer service: Discover Financial Services is tapping the tech to reduce call center response time.
- Engineering and development: Citizens Bank is using generative AI for coding and code conversion.
- Customer experience: Poland-based Nest Bank is piloting Microsoft’s Open AI for its generative AI chatbot, which receives about 11,000 requests for chat daily.
- Back-office operations: Visa is applying gen AI to its payments ecosystem, including using the tech for risk management and fraud prevention.
While applications have been identified at some institutions, others are considering how to approach the tech, Neeraj Mathur, vice president of solutions engineering at generative AI consultancy firm Kognitos, told BAN. To start, “generally our recommendation really is to pick low complexity, high impact, high value projects,” he said.
To understand how some have embarked on their gen AI journeys, Bank Automation News asked financial leaders from four institutions: What questions are you asking yourself and your institution when approaching generative AI?
- As a bank, what can be done with generative AI?
- Can we use generative AI internally, externally or both?
- Is generative AI reliable enough for use within our financial institution?
- Which infrastructure should we use for generative AI?
- Why do we need an external generative AI provider?
- How can banks ensure that generative AI creates only highly accurate content — considering that even the smallest margin of error could have significant repercussions?
- What do we know, with high confidence, that gen AI is good at today and what problems can this solve for our bank?
- How should possible “hallucinations” be addressed?
- Is our bank responsible for what our gen AI says on our website, in online banking or over the phone? [Hint: yes.]
- How do I test and evaluate the risk of adversarial prompts through engineering — for internal and external use cases?
- Do I want to be a gen AI product “builder” or do I wait until the software vendors I already partner with integrate gen AI into their products?
- How can I foster a culture of experimentation and learning around generative AI internally?
- How can generative AI be used to mitigate financial risks and enhance regulatory compliance?
- How can I test and learn about generative AI while minimizing my exposure?
- What are the biggest benefits and risks of allowing my team to use a ChatGPT-like interface to access internal bank knowledge and data?
- What are the biggest benefits and risks of letting customers interface with the bank through a generative AI interface?
- What are the implementation and operational challenges of moving gen AI from experiments to at-scale production deployments?
- What guardrails must I have in place?
- What’s the difference between AI and generative AI?
- How are other financial institutions using generative AI? Am I behind on implementation?
- How much will generative AI cost?
- How are others gaining value from generative AI?
- How do I get my team ready to use generative AI tools?
- How do I get to a responsible, ethical and defensible generative AI framework?
- What data should I build gen AI on?
Leaders from Amdocs, Deluxe, Glia and Kognitos contributed to this list.
