Building intelligent banking experiences with AI

Consider what it means for your business to have AI-enabled smart digital experiences. In financial services, for example, customers are looking for financial guidance, but the information involved is sensitive. Most customers expect the highest levels of security from financial products. Before they entrust their financial data and decisions to AI, personal trust must be established—especially if they feel contact with a human is complex or difficult.

For an AI-powered tool to truly shine, it should provide simple, meaningful results or services that are valuable enough to replace manual processes. American Express, for instance, introduced Amex Trip Planner last year, which can create an “all-encompassing digital itinerary” for travel based on customer preferences and interests. Cardmembers can filter results and see automatic recommendations.

For banks, GenAI can help deliver better service and new solutions to customers, but security and trust are again primary issues. Reaching human support easily and immediately is still important for consumers. GenAI shouldn’t replace human contact but instead act as a tool to optimize customer support experiences. 

Humans must also define the strategy for what products to build. “AI isn’t a one-size-fits-all solution. Meaningful business transformation will come from knowing when to leverage AI, and when not to. Embedding experimentation into how you release products will ensure that you don’t spend time on things that won’t work and foster the environment to innovate, change direction, and deliver something with real value,” says Heather Ryan, Lead Data Strategist and Shadow Board Member at Kin + Carta.

Businesses should also use data intelligence to inform product development. The main challenge, Ryan notes, is that many products are dictated only by business needs rather than a blended view of business and customer needs.

“Synthetic data can help fintech companies develop better products that focus on customer needs,” says Amarendra Limaye, Strategy & Innovation Lead Consultant at Kin + Carta. Synthetic data is the creation of datasets that mimic real-world data and can be used for testing new products while fully protecting user privacy. This can overcome product testing challenges in financial services where creating test environments with access to complex production data is often difficult and increases the potential for malicious attacks. 

Synthetic data allows companies to generate realistic copies of real-time data for extensive testing. However, traditional methods of creating synthetic data as a mirror of real-world data have security vulnerabilities that can lead back to the original data source. Without explicit remediation, there is also a real risk of creating unintended bias or replicating bias present in data captured. By introducing controls, companies can start to address potential bias in generated data. Although the maturity of AI in generating bank-specific data is uncertain, there are companies claiming to possess this capability.

“The potential is there, and the use of synthetic data is one of the most exciting use cases for GenAI applications in financial services,” Limaye says.

Conduct secure experiments

Customer data isn’t the only challenge when using AI to develop products. The recent surge in fraud attacks related to GenAI has added another layer of complexity.

“Financial services have an understandable anxiety when it comes to customer data. How can we uncover insights without compromising our duty to the customer?” Limaye says.

While GenAI is in its early days, the eagerness of banks to begin experimentation is encouraging. It shows that traditional banking institutions can successfully navigate the ever-changing fintech landscape and deliver cutting-edge experiences to customers.

Adapting to change begins with understanding it, and the most effective way to leverage any new technology is to figure out how it can improve your processes and results. Once that’s clear, you can start experimenting in a sandbox environment, gathering data on what works and what doesn’t. Experimenting with AI is all about taking an iterative approach, gaining insights, and getting ready for the future.

Companies are weighing the potential risks associated with using third-party AI solutions versus building in-house AI systems to ensure data security. In February, JPMorgan restricted employees from using ChatGPT over compliance concerns and restrictions the bank has on using third-party vendors.

The rise of third-party vendors from popular web platforms creates the risk of “shadow AI,” Hilali explains. If employees are using business data with third-party vendors without the awareness of their employer, “you lose control of knowing where your data is,” she says.

Improve your data hygiene

Before embarking on their AI journey, companies should establish a solid foundation for data management by implementing better data hygiene practices. Good data hygiene “can start with simply using the same terminology across the organization,” Limaye says.

This approach requires companies to make sure all departments are using the same terms to describe their technology, processes, and innovations. This can be incredibly difficult, Limaye notes, because many large organizations have been built through acquisitions and have gone through transformations that empower local teams, resulting in a wide range of tools and methodologies. 

More complex data hygiene tasks could involve standardizing the application programming interface (API) standards. For example, in some global companies, divisions in different countries may use different API standards for the same data, which prevents data collection and analysis across boundaries and limits the insights that can be gathered.

“Although the intention to empower your local teams makes sense, a lack of guidelines results in wasted enterprise energy,” Limaye says. “It results in unintentional inefficiencies that could have been avoided, and ultimately, limits your innovation potential.”