Big US Banks Are Finally Counting AI Returns — What Q3 2026 Earnings Teach Every Business

Introduction

For three years, America's biggest banks talked about artificial intelligence in terms of potential: pilots, hours saved, "transformational" opportunities. In October 2026, as the third-quarter earnings season opened, the conversation changed. Banks have started reporting what AI actually costs and what it actually returns.

Bank of America's CEO Brian Moynihan put a number on it: roughly 130 to 140 AI projects that cost about $400 million are generating around $800 million in benefits — and the bank plans to double its AI budget next year. JPMorgan, which has set aside roughly $19.8 billion for technology in 2026, topped the Evident AI Index for Banks for the fifth year in a row. This guide explains what the banks are really getting from AI, where the returns come from, and what US businesses and fintech companies should copy.

The Problem: AI Spending Without Clear Returns

Most organisations have the same experience with AI: lots of activity, unclear payback. Industry research from PYMNTS suggests that while 90–100% of large US enterprises report some return from AI, only around 5–10% report full payback on their investment. Teams can point to time saved, but finance leaders want to see lower costs or higher revenue on the income statement.

That gap is why bank earnings matter. Banks are among the largest, most measured AI spenders in the world. When they start publishing cost-versus-benefit figures, they provide a blueprint for every other business trying to justify its own AI budget.

Why It Matters

Banks move the rest of the market. Their AI decisions shape what fintech vendors build, what regulators expect, and what customers come to see as normal service. If Bank of America's customers can resolve questions instantly through an AI assistant, customers of a regional lender or a fintech app will expect the same.

Investors are also watching closely. Early coverage of the October results showed that strong AI spending does not guarantee a strong quarter, and JPMorgan's CFO reportedly warned that the bank would not be immune if the wider AI build-out went into reverse. The message for every business is the same: AI has to earn its keep in measurable results, not just in headlines.

Where Banks Are Getting AI Returns, Explained

1. Customer service assistants. Bank of America's virtual assistant, Erica, had around 24.6 million active users in the second quarter. The bank says it would need roughly 11,000 additional customer service staff without it, and its agent-assist tool cuts average call times by nearly a minute.

2. Employee productivity. At JPMorgan, around 150,000 employees use the internal LLM Suite every week, with users reporting roughly four hours of time saved per week. More than 90% of engineers in its Commercial & Investment Bank use AI coding assistants.

3. Software development. AI coding tools have lifted developer productivity at Bank of America by more than 20%, while Citigroup's 10,000-plus engineers using advanced AI tools have reportedly freed up around 100,000 hours of weekly capacity.

4. Operations and compliance. JPMorgan reports that AI-driven transaction screening doubled processing volumes while cutting manual checks in half.

Benefits

• Lower cost to serve. AI assistants handle routine questions at a fraction of the cost of a call centre.
• Faster software delivery. AI coding tools help engineering teams ship more with the same headcount.
• Better compliance throughput. Screening and monitoring scale without hiring in proportion.
• Happier customers. Instant answers and shorter calls improve the experience.
• Measurable ROI. When AI projects are tracked like any other investment, budgets can grow with confidence.

Real-World Examples

Large banks: Bank of America has approved around 300 AI use cases and deployed more than 100 across the bank — but it measures them as a portfolio, with a clear total cost and total benefit.

A regional lender: a mid-sized US credit union adds an AI assistant to its online banking that answers balance, card and payment questions, and hands complex cases to staff with a summary already written. Call volume drops and members get answers at 10pm.

A fintech startup: a payments company in New York uses AI to triage support tickets and draft compliance reports, letting a small team serve a fast-growing customer base — the kind of system an AI development company in the USA can build on top of existing tools.

Common Mistakes

• Counting hours, not dollars. "Hours saved" only matters if it turns into lower costs, more capacity or more revenue.
• Too many disconnected pilots. Dozens of small experiments with no shared tracking make ROI impossible to prove.
• Ignoring adoption. JPMorgan's results come from weekly use by tens of thousands of staff, not from tools nobody opens.
• Skipping governance. Banks pair AI with strict controls; firms that don't risk errors, fines and lost trust.
• Assuming the boom lasts forever. Even the biggest spenders warn that AI markets can turn — plan for value that holds up either way.

Best Practices: Copy the Banks

Start with use cases that have a clear before-and-after number: support tickets resolved, call minutes saved, documents processed, hours of manual review removed. Track cost and benefit for every project in one place, and expand only what pays back. Invest in adoption — training and integration into daily tools — because unused AI returns nothing.

Most mid-sized businesses can't build JPMorgan's LLM Suite, but they don't need to. An AI automation agency in the USA can connect AI assistants to your existing CRM, help desk and documents in weeks. For fintech products, a custom software development company in the USA can build AI features with the audit logs and controls financial customers expect. Our free AI Readiness Assessment is a quick way to find your highest-ROI starting point.

Conclusion

October 2026 marks the point where AI in banking moved from promises to profit-and-loss statements. Bank of America's 2-to-1 return, JPMorgan's weekly adoption at massive scale and Citi's engineering gains all point to the same lesson: AI pays when it is focused on measurable work, adopted widely and tracked like any other investment. Businesses and fintech companies across the USA can apply that discipline today, at a scale that fits their size.