Bank of America AI Strategy: Bigger AI Bets, Each with a Payback

Bank of America ranks 7th in the Evident AI Index, up three places, and has run its customer AI assistant, Erica, since 2018. Its strategy backs AI that scales across the whole bank, and every AI project in production has a defined payback. It expects AI benefits to rise from $250 million to $300 million in 2026 to about $700 million in 2027.

The ambition is a “high-tech, high-touch” bank. The bank uses that phrase for its client experience, with Erica as “the bedrock,” according to an August 2025 press release.

AI is also an efficiency engine. At its November 2025 investor day, its first in nearly 15 years, CEO Brian Moynihan said AI could deliver 10% to 15% efficiency gains, to be reinvested in growth or taken to the bottom line.

By October 2026 he said: “We are 10% more efficient, so next year we get 10% more work for the same cost.”

The focus is scale. Chief Technology and Information Officer Hari Gopalkrishnan prioritizes AI that can be used across all eight of the bank’s business lines, according to Fortune. He favors redesigning processes end to end: “You start to pick apart that process and reimagine it. That’s when you get ROI.” That kind of rebuild depends on Workflow Literacy: knowing a process well enough to redesign it.

It says what it will not prioritize. Gopalkrishnan is less interested in AI tools that save a few minutes on simple tasks. Bank of America is the first company in this index to state a deprioritization publicly.

AI is deployed by role. The bank’s main tools are:

• clients: Erica, the in-app assistant
• service representatives: EricaAssist, with generative AI since July 2026
• advisors: an AI meeting tool for Merrill and Private Bank advisors (March 2026)
• commercial bankers: a generative AI tool for prospecting and onboarding (September 2026)
• developers: AI coding tools
• all employees: Erica for Employees, for IT and HR questions

The budget is large and rising. Bank of America spends about $13 billion a year on technology, with about $4 billion for new initiatives, including AI. Moynihan said the bank had spent about $250 million on new AI deployments in 2026 by June, as reported by American Banker.

In September 2026, the bank said it plans to double its AI expense budget next year, according to a conference summary.

Erica is narrow on purpose. It is a constrained model connected to 110 systems and about 700 customer intents. Moynihan explains why: “the answer has to be perfect if you’re in our business and trusted.”

Erica is becoming an agent. On the Q3 2026 earnings call, Moynihan said Erica can now complete tasks on an agentic basis. A human must still approve any money movement: “We believe strongly in ‘human in the loop’,” as reported by AIM Media House.

Access is universal. In November 2025, Moynihan said all 213,000 employees were getting access to AI. Training covers AI basics and advanced prompt engineering. About 200,000 employees use AI in their daily work, mostly Erica for Employees (June 2026).

The bank has not disclosed a usage mandate or a required training standard.

The workforce message has shifted. At the investor day, Moynihan said: “I don’t really want to say to people, ‘I can see this taking our overall headcount down.’”

In January 2026, he said AI coding tools were doing the work of about 2,000 developers and that he expected headcount to fall in 2026, as reported by The Stack. In April 2026, he credited technology with eliminating 1,000 positions through attrition, according to Entrepreneur.

AI is also creating work. In Q2 2026, the bank hired about 1,000 contractors to patch software flaws found by Anthropic’s Mythos model, alongside about 3,000 technical staff, as reported by The Register.

Ownership sits with technology leadership. Gopalkrishnan, CTIO since July 2025, oversees the strategic technology budget. No Chief AI Officer has been named.

Governance is described as controls, not published as a framework. Human approval for money movement and a deliberately constrained Erica are the main public controls. The sources reviewed show no published responsible AI principles. Evident ranks the bank 13th on Transparency, which scores responsible AI activity.

AI fluency, coded against the Lead with AI Fluency Matrix (September 2026 edition):

• Fluency floor: Not disclosed. Every employee has access and training is offered, but no mandate or standard was found.
• Highest level evidenced: Level 3 (Integrating), company-claimed. AI coding tools gave 18,000 developers a 20% productivity lift in selected stages of development (November 2025), and Erica for Employees has cut IT help desk calls by more than half. Erica now works as an agent with human approval for money movement, but no results for that mode have been published, so Level 4 is close but not yet evidenced.
• Portfolio design: Multi-level, by role. Clients, service representatives, advisors, commercial bankers, developers and all employees each have their own tools. No role-specific fluency targets have been disclosed.

Every AI project needs a payback. On the Q3 2026 call, the bank said it has about 130 AI projects in production, each with a defined payback. All figures below are company-claimed.

It reports AI value in money:

• AI benefits of $250 million to $300 million in 2026
• about $700 million expected in 2027, from projects already in place
• benefits of two to three times the expense over time, in Moynihan’s words

A broader figure includes automation. In September 2026, the bank put the benefit of AI and automation together at $800 million, on $400 million of cost.

Erica has the longest track record. In 2025, Erica had 20.6 million users and nearly 700 million interactions, passing 3.2 billion interactions since 2018, according to the bank’s March 2026 digital report, as summarized by CMSWire.

Employee tools report time saved:

• Erica for Employees has cut IT help desk calls by more than half.
• AI coding tools gave developers a 20% productivity lift in selected stages.
• The advisor meeting tool can save up to four hours per meeting.

The roadmap runs to 2027. The AI budget doubles next year, and benefits are expected to reach about $700 million. Moynihan has said Erica will eventually move to open-ended agentic AI, without a date.

Change log

• August 2025: Erica passes 3 billion client interactions.
• November 2025: The first investor day in nearly 15 years is held, and all 213,000 employees get AI access.
• January 2026: Moynihan says AI coding tools do the work of about 2,000 developers and expects headcount to fall.
• March 2026: An AI meeting tool launches for Merrill and Private Bank advisors.
• April 2026: Moynihan credits technology with eliminating 1,000 positions through attrition.
• Q2 2026: About 1,000 contractors are hired to patch software flaws found by Anthropic’s Mythos.
• July 2026: EricaAssist gains generative AI for more than 18,000 service representatives.
• September 2026: A generative AI tool for commercial bankers launches, and the bank plans to double its AI budget next year.
• October 2026: Evident ranks the bank 7th, up three places, and the Q3 call reports about 130 AI projects, each with a defined payback.

• Disclosure: 4 of 4. Bank of America answers all four questions, including a stated deprioritization and a dated value target. Governance is its thinnest answer.
• It has the clearest focus rule in the index: back AI that scales across the bank, redesign processes end to end and skip small time-savers.
• It treats AI as an investment. Every project has a payback, and the bank publishes the total.
• The say-do gap is the workforce message. In late 2025, Moynihan did not want to talk about lower headcount. In 2026, the bank cut roles through attrition.
• Its weakest Evident pillar is Leadership (17th), which scores how a bank communicates its AI strategy, while it ranks 7th on Talent and 8th on Innovation.
• Watch 2027. The test is whether benefits reach about $700 million as the AI budget doubles, and how Erica performs as an agent.

What the numbers show: On a comparable basis, revenue rose 10% from 2023 to 2025 while headcount held at about 213,000. That is the pattern its CEO describes, with technology absorbing roles through attrition, but the bank has not put a number on AI’s share.

Revenue, headcount and revenue per employee, FY2022 to FY2025

• FY2022: $94.95bn revenue, 217,000 employees, $438k per employee
• FY2023: $102.8bn revenue, 213,000 employees, $482k per employee
• FY2024: $105.9bn revenue, 213,000 employees, $497k per employee
• FY2025: $113.1bn revenue, 213,000 employees, $531k per employee

Change from FY2023, the first comparable year, to FY2025: revenue up 10%, headcount flat, revenue per employee up 10%.

Notes: Revenue is total revenue, net of interest expense, as reported. Headcount is approximate, at year-end. In its 2025 annual report, Bank of America restated 2023 and 2024 revenue to a new presentation, about $4 billion a year higher. 2022 is on the old basis, so the comparable trend starts in FY2023.

These figures give context on scale and productivity. They do not show that AI caused any change. Sources: Bank of America 10-K filings.

What is Bank of America’s AI strategy?

Bank of America backs AI that can scale across all eight of its business lines and favors redesigning whole processes over small time-savers. Every AI project in production has a defined payback, and the bank expects AI benefits of about $700 million in 2027.

‍

What is Erica?

Erica is Bank of America’s AI assistant for clients, launched in 2018. It had 20.6 million users in 2025 and has handled more than 3 billion interactions. The bank says Erica can now complete some tasks as an agent, with human approval required for money movement.

‍

How much does Bank of America spend on AI?

The bank spends about $13 billion a year on technology, with about $4 billion on new initiatives including AI. Moynihan said it had spent about $250 million on new AI deployments by June 2026, and the bank plans to double its AI budget next year.

‍

Is Bank of America cutting jobs because of AI?

Bank of America has not announced AI-related layoffs. Its CEO credited technology with eliminating 1,000 positions through attrition in 2026 and said he expects headcount to fall.

‍

How is this profile put together?

Lead with AI profiles use public sources only, such as annual reports, earnings materials, filings and on-record interviews, dated at the time of writing. AI fluency is coded against the Lead with AI Fluency Matrix. Company figures are marked as company-claimed unless they have been independently verified.