By PYMNTS | October 22, 2026
Highlights
Many banks still run cores up to 40 years old, and full conversions can take years and cost millions. Sutherland and FlowX.AI will run 220-plus prebuilt agents on top of those systems instead.
FlowX.AI says one European lender cut manual credit handoffs by 80%, and another institution cut lending operating costs by 40%. It did not name either one.
Kastle CEO Rishi Choudhary calls agents an “action layer” over legacy cores. Banks widen agents’ access only as performance warrants. A collections agent, for example, can process a payment but cannot reverse it. Kastle raised $24 million in September to work on top of existing cores.
Banks keep their core systems because replacing them is costly and disruptive. Keeping them carries its own price. Employees work across fragmented applications, repeat manual steps and rely on expensive integrations.
Fifty-two percent of financial institutions name legacy technology as an obstacle to real-time payments modernization, and 53% name manual-intensive internal processes, PYMNTS Intelligence found. Many banks and credit unions still run legacy core systems up to 40 years old, Federal Reserve Bank of Kansas City researchers wrote in a 2024 briefing. Full conversions can take several years and cost millions, they said.
Sutherland and FlowX.AI are betting banks don’t have to choose. Their new partnership puts AI agents to work on top of the cores banks already run.
Sutherland, FlowX.AI Put 220-Plus Agents on Existing Bank Systems
Sutherland, an artificial intelligence-driven business transformation company, will deploy and run the AI agents that FlowX.AI supplies. FlowX.AI, a platform for building, deploying and governing enterprise AI agents, offers more than 220 prebuilt agents. The companies announced the partnership Monday (Oct. 19). The deal covers banking, insurance and other regulated industries in North America and Europe.
Under the partnership, Sutherland defines the roadmap and business case, redesigns manual workflows around agents and people, and configures or builds the agents. It connects them to a client’s existing applications, data and core systems. It then runs and monitors them against agreed business outcomes, the companies said.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
In banking, the agents will target onboarding, account opening, lending, mortgage underwriting, anti-money laundering checks and customer retention. FlowX.AI said it ships preassembled production stacks for retail mortgage underwriting and commercial onboarding.
“Regulated industries hold the biggest opportunity for AI, and most of it is still stuck in pilots,” FlowX.AI CEO Ioan Iacob said in the announcement. FlowX.AI pointed to results from two earlier deployments. A European commercial lender removed 80% of manual handoffs in its credit workflows, the company said. Another institution serving more than 4 million customers cut lending operating costs by 40%. FlowX.AI did not name either institution.
AI Startups Split Between Building on Bank Cores and Replacing Them
Kastle, an agentic AI platform for consumer lenders, takes a similar view. Agents act as an “action layer” on top of legacy cores, CEO Rishi Choudhary told PYMNTS CEO Karen Webster in October. Banks start agents on narrow, repetitive tasks and widen access as performance warrants, Choudhary said. A collections agent can process a payment but cannot reverse it. That request goes to a person or another authorized agent. The gap between a model demo and a regulated bank is wide, he added.
AI agents can reach those cores through APIs, Model Context Protocol (MCP) connections or computer use, PYMNTS reported in September. Computer use lets an agent operate software through the same screens an employee uses. Mambu, a core banking provider, offers an MCP connection. It lets agents search and run more than 300 operations in its APIs, and banks set which operations each agent may use, the company said in May.
Startups building for those workflows are drawing fresh funding. Kastle raised a $24 million Series A led by Insight Partners in September, with participation from Y Combinator and Commerce Ventures.
Maximum is betting on replacement instead. It raised a $30 million seed round led by CRV in August to replace the decades-old infrastructure behind most banks with an AI-native operating system. Cloud core provider 10x Banking is making the same bet. The company, which sells core banking software to banks including Westpac and Chase UK, secured £40 million in debt and equity from AshGrove Capital in August.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.