How Bank of America Built Erica AI to 3.2 Billion Interactions & 24.6 Million Active Users

98% of people who ask Erica a question get an answer without ever talking to a human. That's the whole story in one stat.

Bank of America built Erica back in 2018. Most people wrote it off as another banking chatbot destined to annoy customers and get ignored. Seven years later, here's where it stands:

• 24.6 million active users as of Q2 2026, up 23% year over year.
• 200 million customer interactions in a single quarter, up roughly 15% from last year.
• Since launch, Erica has crossed 3.2 billion total client interactions, now averaging more than 58 million conversations a month.
• The average interaction takes just 48 seconds. Clients get their answer and move on with their day.

Deterministic AI vs. Generative AI in High-Stakes Finance

Here's the part most companies rushing into AI right now completely skip.

Bank of America built Erica on deterministic AI, not generative AI. In a category where a single hallucinated account balance destroys customer trust forever, they chose predictability over flexibility.

Head of Digital Platforms Jorge Camargo put it simply: Speed and accuracy build trust. Trust builds repeat usage. Repeat usage is the entire product.

And it shows up in the business numbers. Digitally enabled sales now make up 70% of total sales at the bank, up from 65% a year ago. Nearly 80% of consumer households are active digital users.

Build the Trust Layer Once. Reuse It Everywhere.

Erica isn't just a chatbot anymore either. It now powers Ask Merrill, Ask Private Bank, and CashPro Chat - the same underlying AI reused across completely different business lines instead of building a new tool from scratch every time.

That's the real product lesson here.

Build the trust layer once. Reuse it everywhere. Most companies rebuild their AI from zero for every new use case and wonder why nothing scales.