UK-based core banking engine provider SaaScada has published research identifying a significant gap between UK banks' agentic AI ambitions and their operational readiness to deploy it. The report, Who's Ready for Agentic AI in Banking?, is based on a survey of 150 UK banking innovation leaders conducted in March 2026, covering retail and business banks with balance sheets of between GBP 0.5 billion and GBP 100 billion.
The findings show that 91% of respondents believe agentic AI will enable entirely new ways of designing banking services. Yet only 31% are actively deploying any type of AI in core operational or decision-making processes, and just one in ten banks fully automates key operational tasks.
Legacy infrastructure and data quality as primary barriers
The research identifies three principal constraints on AI adoption. 77% of respondents cite legacy systems restricting data availability as significantly limiting their ability to adopt AI, with an equal proportion pointing to poor data quality, and 71% citing difficulty accessing real-time data. These infrastructure constraints underpin the low level of current deployment despite high strategic intent.
Furthermore, reliance on manual processes remains pervasive. Between 37% and 42% of banks still rely heavily on manual processes and exception handling for tasks including processing standing orders, scheduled payments, and direct debits, daily interest accrual and posting, account maturity instructions, and scheduled interest rate changes. Only 10% to 13% of banks fully automate these functions. Overall, 61% of respondents describe such tasks as very or extremely painful in terms of cost, manual effort, and risk.
The correlation between automation and perceived operational complexity is direct: 85% of those with minimal automation describe these processes as very or extremely painful, compared with 55% for those with some automation and regular manual oversight, and zero for those who fully automate.
Regulatory explainability and financial inclusion risks
The research also surfaces concerns about the regulatory and social implications of deploying AI on inadequate data foundations. 79% of respondents believe that without high-quality, explainable data, AI could worsen financial exclusion rather than improve it. Despite that concern, only 12% are very confident their organisation could clearly explain and justify AI-driven decisions to regulators today.
SaaScada's chief technology officer, Paul Payne, noted that banks could not expect to innovate with agentic AI while still dependent on manual processes. Payne added that the priority has to be maturing the infrastructure and driving automation first. Only afterwards can banks layer in AI and begin to see real operational benefits.
Industry context
The research reflects a structural challenge facing a significant portion of the UK banking sector. While agentic AI has attracted widespread attention as a transformative capability, its effective deployment in financial services requires a level of data quality, real-time accessibility, and process automation that most institutions have not yet achieved. The findings suggest that for many banks, the path to agentic AI runs through core infrastructure modernisation rather than through AI deployment itself, a sequencing challenge that may delay practical benefits for institutions that attempt to build advanced capabilities on unmodernised foundations.
