Artificial intelligence is beginning to attack one of banking’s oldest and most expensive technological problems: decades-old legacy software, including COBOL systems that still underpin core functions at major financial institutions. New AI coding agents could dramatically reduce the time and expense required to understand, refactor, test and eventually replace portions of those systems, with Deloitte estimating AI could reduce banking-industry software investment costs by 20 to 40 percent by 2028. Yet the larger economic question is whether banks will actually retain those savings. Competition for loans could force institutions to surrender some technology savings through narrower lending spreads, while AI-powered personal financial agents could simultaneously make consumers more aggressive about moving low-yield deposits into higher-paying accounts and investments. The result could be a striking reversal: AI makes banks cheaper to operate while also weakening the inertia that has historically allowed them to retain inexpensive deposits, potentially transferring a meaningful share of AI’s economic benefits from financial institutions to borrowers, depositors and technology vendors.
Key Takeaways
- AI coding tools are increasingly capable of analyzing, documenting, refactoring and testing legacy banking software, potentially reducing technology expenses and accelerating modernization projects that traditionally required years of expensive engineering work.
- The financial benefits may not remain with banks because competitive lending markets can pressure institutions to pass lower operating costs to borrowers, while AI financial agents could make it considerably easier for depositors to seek higher returns elsewhere.
- Large banks that own significant portions of their technology infrastructure could capture savings directly, while many community banks depend on outside core-processing vendors, meaning the immediate financial beneficiary of cheaper AI-assisted software development may sometimes be the technology provider rather than the bank.
In-Depth
Artificial intelligence may finally give banks a practical weapon against technology accumulated over generations. Legacy COBOL systems remain embedded throughout financial infrastructure, handling critical functions that institutions have historically been reluctant to replace because modernization is expensive, complicated and potentially catastrophic when migrations fail.
AI changes the economics. Coding agents can document unfamiliar software, automate portions of code conversion, assist data migration and accelerate testing. Deloitte estimates AI could eventually reduce banking software investments by 20 to 40 percent, while Citigroup is already using AI to automate coding, migrate legacy data and speed testing.
But reducing expenses does not guarantee that shareholders keep the savings. Lending is intensely competitive. When multiple institutions pursue the same borrower, lower technology costs can translate into lower loan pricing rather than permanently higher margins.
Deposits create an even more consequential challenge. Banks have historically benefited from customer inertia: consumers routinely leave money in low-yield accounts because researching alternatives, opening accounts and transferring funds requires effort. Agentic AI could steadily eliminate that friction by monitoring rates and identifying better alternatives.
That creates a fascinating free-market counterweight to AI-driven bank efficiency. The same technological revolution that makes institutions cheaper to operate could make their customers substantially more economically rational. Large banks may save billions modernizing software, yet be forced to share those gains through cheaper credit and better deposit rates. AI therefore may not merely modernize banking infrastructure. It could redistribute economic power toward consumers by making competition faster, easier and considerably harder for financial institutions to avoid.
Sources
- https://www.zerohedge.com/ai/who-keeps-money-when-ai-rewrites-bank-code
- https://www.reuters.com/business/finance/citigroup-says-ai-helps-speed-account-openings-systems-upgrades-2026-04-08/
- https://www.deloitte.com/us/en/insights/industry/financial-services/financial-services-industry-predictions/2025/ai-and-bank-software-development.html
- https://www.pymnts.com/news/artificial-intelligence/2026/ai-expands-banks-options-for-modernizing-legacy-cores/
- https://www.deloitte.com/us/en/insights/industry/financial-services/future-of-software-engineering-in-banks.html