400,000 Daily Prompts Put Bank of America AI on Trial

Can Bank of America AI turn 400,000 employee prompts a day into durable profit gains, or is this still too early to separate real efficiency from earnings-call gloss?

That is the question underneath Brian Moynihan’s second-quarter message. Bank of America executives used the quarter to frame AI as an active operating tool, not a lab project, according to American Banker. They tied it to revenue growth, productivity gains, stronger digital engagement, and better client service.

Moynihan said the bank has approved more than 300 AI use cases, “all of which have good economics.” That phrase matters. Banks don’t get paid for novelty. They get paid for lower costs, faster decisions, better risk control, and more productive client coverage.

“Our bankers automate the research and presentation materials,” Moynihan said. “Our developers code more efficiently and all our teammates improve productivity, consistency, and client service while creating significant opportunities ahead of us.”

Can Bank of America AI make earnings efficiency more than a slogan?

Moynihan’s pitch was direct: AI is already inside the bank’s workday.

Employees are generating more than 400,000 prompts a day. The bank has approved more than 300 AI use cases. 19,000 developers are using AI for real-time coding assistance, which the bank said is increasing productivity by more than 20%.

That is the strongest part of the case. Developer productivity is easier to measure than vague claims about “transformation.” If the same technology budget produces more usable code, the economics become visible inside project timelines, maintenance cycles, and product delivery.

Moynihan put it plainly:

“So the same amount of money in ’27 will get us more code, for lack of a better term, in ’28,” Moynihan said. “So we’re driving everything as hard as we can.”

The harder issue is attribution. Bank of America reported solid second-quarter consumer banking results, but the source material does not prove AI caused all of them. Management linked AI to better service, productivity, and employee focus. Other forces may also sit behind the numbers, including customer activity, cost discipline, product mix, and broader business conditions.

That distinction matters for investors. Bank of America AI is credible as an earnings story only if it keeps showing up in repeatable metrics, not just executive language.

Which Q2 numbers actually support Moynihan’s “good economics” claim?

The second-quarter figures give management something real to point at.

In consumer banking, net income rose 10% year over year to $3.3 billion. Revenue climbed 5% to $11.3 billion. CFO Alastair Borthwick said the segment produced its fifth consecutive quarter of positive operating leverage, held a 51% efficiency ratio, and delivered a 29% return on allocated capital.

Deposits rose to $957 billion. The bank opened 162,000 net new checking accounts. Card spending increased 9% year over year to $266 billion. Consumer investment assets reached $640 billion, up 18% year over year.

Digital metrics were also central to the AI argument:

• Active digital users: roughly 50 million
• Active Erica users: more than 24 million
• Digital sales: 70% of total sales
• Approved AI use cases: more than 300
• Employee prompts: more than 400,000 a day
• Developers using AI coding help: 19,000
• Developer productivity gain cited: more than 20%

Borthwick tied those digital and AI tools to customer and employee outcomes.

“Digital engagement remains a clear differentiator, with roughly 50 million active digital users, more than 24 million active Erica users and digital sales representing 70% of total sales,” Borthwick said. “New AI capabilities have improved service, increased efficiency and allowed teammates to focus on higher-value client interactions.”

XOOMAR analysis: In banking, “good economics” should mean a lower cost per interaction, faster preparation for client meetings, fewer manual steps, better software output per developer, and higher conversion from digital engagement. Bank of America gave evidence for several of those. It did not disclose a clean AI savings number, nor did it isolate AI’s impact from the rest of the quarter.

For readers tracking how bank earnings narratives are shifting around cost and revenue pressure, see XOOMAR’s related coverage of Higher Costs Stalk PNC 2026 Guidance as Revenue Rises and Bankruptcy Spike Jolts 2Q Bank Earnings Credit Nerves.

Where does Bank of America say AI is already changing work?

The clearest disclosed uses sit in four places: client preparation, software development, consumer digital service, and wealth management.

Customer service starts with Erica, Bank of America’s virtual assistant for consumers, launched in 2018. American Banker reports that Erica uses machine learning and natural language understanding to interpret customer questions and map them to prewritten answers in a curated library.

Wealth management has its own version. Merrill Lynch financial advisors use Ask Merrill to prepare for client calls. Borthwick said advisors were more productive during the quarter because of “growing digital engagement and new AI-enabled tools that help advisers prepare for client conversations, identify opportunities and deliver more personalized advice at scale.”

Developer work is the cleanest productivity example. The bank said 19,000 developers use AI for real-time coding assistance, lifting productivity more than 20%.

Corporate and investment banking has a different AI angle. Borthwick said clients are investing in AI, and Bank of America is active in investment banking and global markets around “capital raising” and financing the “massive capital investment and infrastructure build around the world.”

Here is the useful split:

That last row is important. The outline of a bank’s AI future often includes fraud monitoring, compliance workflows, cash-management tools, and call-center assistance. This earnings account did not provide enough detail to treat those as Bank of America disclosures.

Why don’t executives, employees, regulators, and investors hear the same AI pitch?

Executives hear margin defense. Moynihan and Borthwick described AI as a way to improve productivity, consistency, service quality, and organic growth without presenting it as a blunt cost-cutting campaign.

Employees hear something more complicated. Moynihan has said in supplied interview excerpts that Bank of America workers “shouldn’t be worried” about AI as a job threat. Still, the economics he described depend on people doing more with better tools. That can reduce drudge work. It can also change what the bank values: judgment, relationship management, and exception handling gain importance, while routine preparation and processing become easier to automate.

Regulators and risk managers hear a control problem. Moynihan addressed the risk of AI use directly when discussing companies in the bank’s lending portfolios.

“We talk to the companies that are in our portfolios of lending to make sure they’re active using this so they don’t get left behind,” he said. “But on the other hand, they’re using [AI] in a responsible way so that they can protect their data and their security and things like that.”

Investors hear a proof test. They want AI spending to show up in productivity, efficiency, revenue quality, or customer engagement. They don’t want another multiyear technology bill with benefits too vague to audit.

Is this another banking automation cycle, or is generative AI reaching deeper?

The source gives one clear historical marker: Erica launched in 2018. That matters because Bank of America’s current AI message is not starting from zero. The bank has already had years to train customers to interact with a virtual assistant.

What has changed is the spread of AI from customer self-service into knowledge work. Moynihan cited bankers automating research and presentation materials. Borthwick cited advisors preparing for client conversations. Developers are using AI while coding.

That reaches into work banks traditionally treated as professional judgment support, not simple process automation.

XOOMAR analysis: This is where generative AI differs most from earlier digital banking tools. The point is not only to move a customer from a branch to an app. It is to change the labor behind advice, sales preparation, software delivery, and internal analysis.

The warning is just as important. Technology gains rarely flow evenly through an income statement in one quarter. Bank of America has disclosed usage and productivity signals. It has not disclosed a full AI profit bridge.

Who gets squeezed if Bank of America AI keeps scaling?

Customers could see faster answers, more tailored digital interactions, and better-prepared bankers or advisors. The tradeoff is trust. If AI-assisted service produces errors, weak escalation, or unclear data use, speed will not be enough.

Employees face a sharper shift. Routine research, drafting, coding support, and preparation work will likely become AI-supervised workflows. The human premium moves toward client judgment, review, escalation, and accountability.

Rival lenders face the scale problem. Bank of America AI can be spread across roughly 50 million active digital users, more than 24 million active Erica users, and large internal developer and advisor groups. Smaller institutions may need narrower deployments or external tools, but this source does not provide enough detail to compare their cost structures.

For the broader banking sector, the test is simple: AI will be judged by repeatable margin improvement and service gains, not conference-stage enthusiasm.

Which proof points decide whether AI spending passes the next earnings test?

The next phase will come down to disclosure.

Bank of America has already put several markers in public view: daily prompts, approved use cases, active Erica users, developer adoption, and developer productivity gains. If management keeps expanding those metrics, investors will have a better way to judge whether AI is producing operating discipline rather than just absorbing budget.

The strongest near-term evidence should come from internal workflows, customer service, software development, and advisor preparation, because those are the areas Bank of America actually discussed. Fully autonomous financial advice, AI-driven credit decisions, or broad compliance automation were not established in the source material.

The cleanest confirmation would be a pattern: rising digital engagement, stable or improving efficiency, measurable employee output gains, and management continuing to connect AI to revenue and expense discipline without overstating causation.

The weakening signal would be vagueness. If future calls keep the AI rhetoric but drop the productivity numbers, the “good economics” claim gets harder to trust. If Bank of America keeps converting AI usage into visible revenue growth and expense control, rival banks will face the question Moynihan has now put on the table: where, exactly, are their own AI gains showing up?

Disclaimer: This XOOMAR analysis is for informational and educational purposes only. It is not financial, investment, legal, tax, or professional advice. It does not provide buy, sell, hold, price-target, portfolio, or personalized recommendations. Verify information independently and consult qualified professionals before making decisions.

The Bottom Line

• Bank of America is positioning AI as a measurable productivity tool rather than an experimental technology.
• The bank’s reported 400,000 daily prompts and 300 approved use cases suggest AI is already embedded in employee workflows.
• A productivity gain of more than 20% among 19,000 developers could improve technology delivery and long-term operating efficiency.