Bank of America Adds Generative AI to EricaAssist
Bank of America added generative AI to EricaAssist, the real-time support tool used by more than 18,000 customer service representatives. The bank says the system can deliver contextual guidance in under three seconds and has reduced average call time by nearly one minute, while employees retain responsibility for client decisions.
Bank of America announced on July 21 that it has added generative AI capabilities to EricaAssist, its employee-facing support system for customer-service calls. The bank says more than 18,000 representatives use the tool, which surfaces information and suggested next steps while the employee remains responsible for the conversation and decision-making.
The new capabilities can summarize why a client is calling, collect relevant account and service information, and recommend next steps based on the employee's role and the client's relationship with the bank. Bank of America says that guidance appears in under three seconds and that EricaAssist has reduced average call time by nearly one minute. Those performance figures are company-reported; the announcement does not provide a measurement period or independent validation.
What changed in the workflow
EricaAssist is designed as an internal assistant rather than a consumer chatbot. It operates during a live conversation, giving the representative context without requiring the client to leave the call or interact directly with a model. Bank of America plans to expand the system to additional servicing scenarios and business lines later in 2026.
The bank also says it spends $14 billion annually on technology, with more than $4 billion allocated to new initiatives including AI. That budget gives the deployment scale, but it should not be read as evidence that every AI initiative produces the same operational benefit.
Why the human role matters
The material design choice is the division of responsibility: the system summarizes and recommends, while the employee listens, explains options and decides how to respond. That human-in-the-loop structure can reduce the risk of an unsupported suggestion reaching a client unchecked, although it does not remove the need for access controls, audit logs, quality monitoring and escalation paths.
For data and operations teams, the next useful evidence will be error rates, escalation frequency and customer outcomes across different call types. For now, the announcement demonstrates a large employee-facing generative-AI deployment and reports faster service; it does not yet establish independently measured gains across the bank's full customer-service operation.
Key Points
- 1Bank of America says more than 18,000 customer-service representatives use EricaAssist during live client conversations.
- 2The new generative-AI features deliver guidance in under three seconds and have reduced average call time by nearly one minute, according to the bank.
- 3EricaAssist remains employee-facing: it summarizes and recommends next steps, while representatives retain responsibility for client decisions.
Scoring Rationale
Bank-scale deployment across more than 18,000 service representatives makes this operationally relevant, while the efficiency figures remain company-reported and the system is limited to employee assistance.
Sources
Primary source and supporting public references used for this report.
Practice with real Banking data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Banking problems
