“We Can’t Afford Any Mistakes,” Says Bank of America Head of Digital Nikki Katz
Why accuracy, governance, and human oversight shape how Erica is deployed across the bank’s AI systems.
“We have 59 million clients that access their bank accounts, their loans, retirement, et cetera, through our digital channels,” said Nikki Katz, Head of Digital at Bank of America on a recent episode of The Buzz podcast. “My team is responsible for ensuring those digital channels are meeting our client needs.”
Bank of America has embedded artificial intelligence into its core digital systems. Erica, its AI-powered virtual assistant, serves as a primary interface linking customer interactions with back-end services and workflows. Launched in 2018, Erica is a central part of the bank’s digital experience.
Erica has assisted nearly 50 million clients and recorded more than 3 billion interactions since launch. Those interactions take place across mobile and web channels and encompass tasks ranging from account management to proactive financial insights.
Erica as a Shared Digital System
Katz describes Erica as an active presence in how clients manage their finances. “Erica really acts, as we call it, a guide by your side for our clients,” she said. The tool helps clients complete tasks and access account information, and it provides information about spending patterns, balance status, and financial obligations.
Erica’s capabilities extend across multiple products, including banking, investments, and retirement services. The system also supports CashPro Chat, a virtual assistant for business clients that provides access to treasury and payment functions. Press releases from Bank of America mention that AI tools in CashPro have helped corporate clients with tasks including forecasting and operational workflows.
The system adapts to client needs as context changes. Katz described how Erica was updated to interpret user queries related to wildfires, helping clients find relevant financial support programs and balance information. This illustrates how the system is updated based on real-time usage patterns.
Internally, Bank of America has deployed AI assistants for employees. More than 90 percent of staff use an internal assistant for tasks such as IT troubleshooting and internal navigation, and this deployment has reduced internal service-desk calls by more than half.
AI models extend beyond chat functions. Bank of America operates hundreds of AI and machine-learning models across fraud detection, marketing, underwriting, and internal productivity workflows. The bank’s technology budget totals approximately $13 billion annually, with about $4 billion allocated to new initiatives such as AI.
Where Bank of America Draws the Line on AI
Katz said the bank designs AI systems around client needs and regulatory standards. “Our business is a trust-based business,” she said. The bank emphasizes accuracy in financial transactions and oversight where needed.
Katz pointed out the need for precision in customer interactions. “If I were to say, ‘Hey Erica, I need you to send $20,’ and Erica had a hallucination, that would not be very welcome,” she said. This is part of an approach where predictive analytics and supervised models dominate interfaces that affect financial decisions.
Generative AI systems are used in workflows that include human review. For example, AI tools now prepare summaries and briefing documents for bankers, helping them manage client meetings more efficiently. “That shaves hours for our bankers and allows them to come really well prepared,” Katz said. Independent reporting confirms that AI has helped bankers manage more clients by automating preparatory work.
Bank of America’s leadership describes this approach as “augmented intelligence.” AI supports employee workflows and client service while incorporating governance, monitoring, and data quality controls.
The impact of AI is visible in operational outcomes. Industry reporting indicates that Bank of America operates numerous machine-learning models that have contributed to reduced fraud and decreased service call volumes.
Katz said the bank continues to expand Erica’s capabilities. “We’re continuing to add more insights, more areas where Erica can help,” she said. Current efforts focus on relevance, timeliness, and simplicity for clients.
“We make over 1,000 different enhancements to our digital experience every year,” Katz said, pointing to the continuous development of AI as core infrastructure.