Manual Operations Stalling Banking Progress as Agentic AI Ambitions Meet Reality

The financial services sector is currently captivated by the promise of agentic AI. Unlike generative AI, which primarily synthesises and creates content, agentic systems possess the autonomy to reason, make complex decisions, and execute multi-step workflows without constant human intervention. For banks on both sides of the Atlantic, this technology represents the next frontier of operational efficiency and risk mitigation.

However, a stark operational reality check has arrived. A breakthrough research from cloud-native core banking engine SaaScada reports a widening gulf exists banks’ architectural ambitions and their execution capabilities. The report, Who’s Ready for Agentic AI in Banking? underscores a fundamental truth: financial institutions cannot automate the future using the manual workflows of the past.

The Ambition vs. Reality Gap

SaaScada’s survey of 150 UK banking innovation leaders, including C-suite executives and digital transformation heads managing balance sheets between £0.5Bn and £100Bn, reveals a profound paradox:

• 91% of innovation leaders believe agentic AI will enable entirely new ways of designing banking services.

• Yet, only 31% are actively deploying any form of AI within their core operational or decision-making processes today.

This execution gap is not born out of a lack of interest, but rather severe infrastructure limitations. Agentic AI requires seamless, real-time access to clean, unified data to function safely and effectively. Instead, legacy infrastructure is choking the life out of innovation.

The data highlights three critical systemic barriers restricting AI adoption, with 77% of innovation leaders citing legacy systems restricting data availability as a primary hurdle. Tied with legacy constraints, poor data quality impacts another 77% of institutions, while 71% point to the ongoing difficulty of accessing real-time data as a significant roadblock to deployment.

As Steve Round, Co-Founder and President at SaaScada, aptly puts it:

“Trying to build AI on ancient legacy foundations is like racing an Aston Martin over cobblestones – it’s going to be a bumpy ride. If banks are serious about getting ahead with AI, they need data and core systems that are fit for purpose. Otherwise, all the ambition in the world won’t translate into results.”

The Manual Operations Problem: High Pain, High Risk

The most damning insight from the research is the banking sector’s continued, heavy reliance on manual processes for bread-and-butter operational tasks. At a time when institutions are conceptualising autonomous AI agents, fewer than nearly one in ten banks have achieved full automation for foundational core banking processes.

Rates of Full Automation across Core Tasks:

• Standing orders, scheduled payments, and direct debits: 10%

• Daily interest accrual and interest posting: 11%

• Account maturity instructions: 13%

• Scheduled interest rate changes: 13%

Conversely, between 37% and 42% of institutions remain heavily reliant on manual workarounds and exception handling to get these tasks across the finish line.

This manual debt comes with a massive operational toll. Overall, 61% of respondents describe these basic processing tasks as “very” or “extremely painful” regarding cost, manual effort, and risk.

Unsurprisingly, SaaScada found a direct correlation between operational pain and a lack of modern automation. Among organisations with minimal automation, 85% find these processes highly painful. For those with partial automation and manual oversight, that figure drops to 55%. Tellingly, for institutions that have achieved full automation, the perceived pain level drops to 0%.

Regulatory Compliance and the Explainability Crisis

For US and UK financial institutions, deploying autonomous agents is a severe regulatory risk. Agencies like the UK’s Financial Conduct Authority (FCA) and the US Consumer Financial Protection Bureau (CFPB) are increasingly demanding strict algorithmic accountability, data lineage, and models that mitigate disparate impact.

SaaScada’s research shows that innovation leaders are acutely aware of these compliance stakes: 79% believe that without high-quality, explainable data, AI implementation could actually worsen financial exclusion rather than improve it.

Yet, despite this awareness, only 12% of respondents feel very confident that their organisation could clearly explain and justify AI-driven decisions to regulators today. If an AI agent denies a loan application, blocks a cross-border stablecoin transaction, or freezes an account, the underlying core banking engine must be able to surface an immutable audit trail of the real-time data points that informed that decision. On legacy architecture, pulling that thread is nearly impossible.

Fix the Plumbing Before Buying the Paint

The business case for agentic AI in front-office functions, such as sophisticated virtual wealth advisors or automated commercial credit underwriting, is undeniable. But SaaScada’s findings serve as a vital warning for fintech and banking operations leaders: AI cannot fix broken plumbing.

If an institution’s core banking engine requires manual exception handling just to post daily interest or process a direct debit, layering a complex, autonomous AI agent on top of it will simply compound operational risk and regulatory exposure.

“Banks can’t expect to innovate with agentic AI if they are still mired in manual processes,” warns Paul Payne, CTO at SaaScada. “The priority has to be maturing the infrastructure and driving automation first. Only then can banks layer in AI and start to see real operational gains.”

To bridge the gap between AI ambition and operational reality, banks must prioritise core modernisation. Migrating away from rigid, siloed legacy systems toward cloud-native, data-driven core engines will naturally eliminate manual operational friction. Once a bank has achieved clean, real-time data flows and 100% automation on basic tasks, it will finally possess the architectural foundation required to satisfy risk regulators and unleash the true power of agentic AI.