Generative AI’s unfolding transformation of banking

Greg Jacobi has worked in banking for more than 20 years in various roles—from leading engineering teams that build digital banking capabilities to managing customer experiences for the enterprise to directing P&L ownership for consumer lending business units. Greg works with banks and lenders to help them align their strategic roadmap to the transformational capabilities that Salesforce can enable to drive positive outcomes and accelerate the mission of their organization.

How can banks benefit from generative AI, and what are the most promising use cases?

Banks tend to take a conservative approach to adopting new technologies because of their strict regulations. However, embracing predictive AI has helped banks become more productive and efficient. Predictive AI is designed to forecast outcomes based on historical data, and its primary goal is to predict future events or behaviors by analyzing patterns and trends.

Generative AI is the next wave of artificial intelligence that can further enhance the banking industry. It involves training a model to generate new data that is similar to the training data it was given. The output is typically some form of content—from text to images to video and even computer code.

Typical use cases include:

• Sales: Auto-generated and personalized emails
• Service: AI-generated service replies, case wrap-up and knowledge article creation
• Marketing: Audience segmentation, campaign creation and personalized outreach

How can banks build trust with their customers?

Banks must decide how to prioritize trust, and working with partners who share the same sentiment about trust is crucial. Generative AI is a rapidly evolving technology, making it imperative that a bank’s partner understands how to leverage customer data without creating additional exposure.

For example, banks can build trust by managing customer data through steps that might include prompt building, secure data retrieval, dynamic grounding, data masking and prompt defense. These steps ensure data privacy and security. After securely processing data with a large language model (LLM), teams should monitor generated responses for toxicity or inappropriate content, and then re-mask to protect sensitive information. Keeping an audit trail for compliance is also critical.

At this stage in generative AI development, we believe in keeping a human in the loop as a necessary guardrail to protect customer trust. Banks must also be transparent with customers about when they are using generative AI to continue to instill trust.

Establishing and protecting customer trust is a top priority. Banks can start with a narrow use case in a low-risk way, such as with a small group of internal experts who can provide feedback on the LLM and its outputs. By establishing this feedback loop with trusted experts, teams can quickly improve the quality of the model’s outputs and optimize its performance over time. Implementing a crawl, walk, run approach with generative AI can help to ensure that a bank’s operations are optimized for benefits and minimized for risk. Banks can establish AI principles to help guide their development and responsible implementation of generative AI.

How can banks prepare and upskill their teams?

Bank employees are expected to have a vast knowledge of regulatory rules and bank policies. Generative AI empowers bank employees with an efficient way to access that regulatory information and automatically generate relevant content based on the situation.

According to the Salesforce Generative AI Snapshot Research Series, nearly 61% of employees who plan to incorporate the technology don’t know how to use trusted data sources or keep sensitive data secure. With technology changing fast, your training practices need to keep up. Start by determining which skills your team needs and set priorities. Make time for employees to build their skills and offer continuous training as generative AI technology evolves.

Take prompt engineering, for example, which is asking specific questions or feeding detailed information into the model to get the best results. Generative AI is creating a new job skill within the bank to effectively engage with these models. Employees who master this skill will yield the most benefit.

Every team member—from the front line to the back office—must know how to engage with and implement these models. Employees who frame a prompt with context and grounding techniques will get the most viable responses, allowing them to provide more effective customer service. Curious employees who experiment with generative AI models might already understand that there is a learning curve to engineering a relevant response.

Recent AI advancements add pressure on data management teams to power algorithms with quality data. According to the Salesforce State of Data and Analytics Report 2023, 86% of analytics and IT leaders agree that AI’s outputs are only as good as its data inputs. It is important to remember that improving data trust takes more than technical investments because organizational culture is critical to driving confidence and adoption. Some 75% of analytics and IT leaders plan to increase investments in training and development.

How will generative AI transform the contact center?

When I worked for a bank, I was surprised by the range of knowledge expected of a contact center agent. One employee could not know everything about all the product and service offerings, the regulatory constraints and the nuances of adjustable-rate and fixed-rate mortgages or layering a CD.

There’s an enormous burden placed on these agents. One real promise of generative AI is the ability to search the entire bank’s knowledge base to help alleviate this burden. Generative AI can also help to ensure that an agent’s responses comply with regulatory requirements, are in the context of the bank’s products and services and are grounded with the customer’s unique data. All of that adds up to massive productivity gains across the business and a higher-quality output.

Effective use of generative AI will impact several key metrics: increasing CSAT (customer satisfaction) scores, decreasing case resolution time and decreasing support costs. A survey we conducted recently for Salesforce Generative AI Snapshot Research Series 2023 found that 90% of service employees currently using generative AI report that it helps them serve their customers faster.

Why jump into generative AI now? Why not wait for broader adoption?

There’s an immediate need to learn and keep up with the pace of generative AI technology. Banks can find areas in an organization where generative AI will produce the most benefit, whether that’s driving banker productivity by augmenting their daily routine, boosting efficiency or improving the customer experience. Banks can’t afford to wait to see how this all plays out amid the breakneck pace of technological change. The only way to stay ahead of the innovation curve is by jumping in now, getting ahead of your skill set and learning how to use generative AI.

The current focus for banks is leveraging generative AI technology to augment human workflows and processes. Yet even with the best digital tools and technologies, there are limitations on the scale of what humans can achieve and the efficiencies gained.

As we observe the direction that generative AI is going in the future, there is potential for more autonomous digital AI agents that can work and collaborate on a bank’s behalf toward a defined goal. These highly specialized AI agents can help banks push past the limits of a human workforce alone and quickly scale business needs.

Here are four tangible recommendations on how banks can get started with generative AI:

• Plan your approach: Start by identifying a business problem, then review your current processes end-toend to identify friction points. As you design a strategy that uses generative AI to improve operational efficiency, emphasize trust as you plan to implement across your bank. Establishing ethical guidelines and guardrails for generative AI use is essential for fostering responsible innovation and creating a secure and trustworthy environment.
• Build a strong data foundation: Given the dependence of AI’s outputs on the quality of underlying data, data management is a high priority. Aligning your data strategy with business objectives is a key starting point, according to the Salesforce State of Data and Analytics Report 2023.
• Enable your people: Get the most out of your AI investment by upskilling your team. Cultivate an AI-ready workforce through the development of specific skill sets, like prompt engineering, and provide employees with the tools and resources needed to enhance their understanding. As generative AI technology evolves, you can offer continuous training to help teams keep pace.
• Ensure trust and security: The power of AI can be undermined by a lack of trust. Update your security and purchasing guidelines to include trust standards like data masking to ensure your data remains safe. Build trust by showing customers you understand pain points, implementing effective data governance and offering transparency.

Greg Jacobi is Vice President and General Manager of Banking Industry Solutions and Strategy for Salesforce.

A version of this article appeared in the BAI Special Report: Leveraging AI With Human Capital. You’ll find more AI insights there.