Banks are adding AI helpers to computer systems up to 40 years old

Many banks still run on core systems that are up to 40 years old. Rather than rip them out, a growing group of companies wants to add AI helpers on top. On Monday, Oct. 19, Sutherland and FlowX.AI announced a partnership to do exactly that, using more than 220 prebuilt AI agents.

What changed

An AI agent is software that can carry out a task, not just answer questions. FlowX.AI supplies the agents. Sutherland will deploy them, connect them to a bank's existing applications and data, and then run and monitor them against agreed business outcomes. The deal covers banking, insurance and other regulated industries in North America and Europe.

In banking, the agents will target onboarding, account opening, lending, mortgage underwriting, anti-money laundering checks and customer retention. FlowX.AI says it ships ready-made setups for retail mortgage underwriting and commercial onboarding.

Why banks keep the old systems

Federal Reserve Bank of Kansas City researchers wrote in a 2024 briefing that full conversions can take several years and cost millions. Staying put has a price too. Employees work across scattered applications, repeat manual steps and rely on expensive connections between systems.

PYMNTS Intelligence found that 52% of financial institutions name legacy technology as an obstacle to modernizing real-time payments. Another 53% point to manual-heavy internal processes.

Where people stay in charge

Kastle, a company that builds AI agents for consumer lenders, calls agents an "action layer" over old cores. Its CEO, Rishi Choudhary, says banks start agents on narrow, repetitive jobs and widen their access only as performance warrants.

• A collections agent can process a payment but cannot reverse it.
• A reversal request goes to a person or another authorized agent.
• Mambu, a core banking provider, lets banks choose which operations each agent may use.

What is claimed and what is not known

FlowX.AI says a European commercial lender removed 80% of manual handoffs in its credit workflows. It says another institution, serving more than 4 million customers, cut lending operating costs by 40%. The company did not name either one, so the results cannot be checked from the announcement.

Not every company agrees with the add-on approach. Maximum raised $30 million in August to replace old bank infrastructure with an AI-native system. 10x Banking, which sells core software to banks including Westpac and Chase UK, secured £40 million that month. Kastle raised a $24 million Series A in September.

What it means for customers

If your bank feels slow, old systems are one reason. Agents aimed at account opening and lending could speed up those steps over time. The material does not say when any bank will roll these out, and no bank customers were named in the announcement.