Banks boost AI surveillance but operational gaps stay

New research from 1LoD, commissioned by communications risk management firm Shield, found that banks are increasing investment in AI-driven surveillance but still failing to deliver broad operational change.

Many banks remain stuck with longstanding surveillance problems despite higher spending on AI. More than 90% of institutions surveyed still operate in alert-heavy environments, while false positives remain the most significant surveillance challenge, with no meaningful improvement from 2024.

The research found that 52% of firms viewed false positives as a highly significant issue. Budget and resourcing constraints followed at 41%, while limited access to quality data and legacy or outdated systems were each cited by 37% of respondents.

Only 7% rated regulatory pressure as highly significant. This suggests the main obstacle for many institutions has shifted from regulatory uncertainty to the practical difficulty of putting new technology into day-to-day use.

The survey drew on responses from senior surveillance and compliance professionals across financial institutions. It compared the latest findings with a 2024 benchmarking survey to assess how the sector's surveillance challenges have changed.

The report describes a widening gap between banks' stated AI ambitions and their ability to put those plans into effect. It identifies legacy technology, fragmented data, and operating models built for earlier market conditions as the main reasons banks struggle to turn investment into measurable change.

Operational gap

Banks continue to face pressure to monitor electronic communications and identify misconduct more accurately. Yet the data suggests many compliance teams are still managing high volumes of alerts that consume time and resources without materially improving risk detection.

Deployment data from Shield, cited in the report and based on evaluations covering millions of communications, showed roughly a threefold reduction in alert noise compared with legacy systems. In direct comparisons, that corresponded to up to 44% higher accuracy and about three times more actionable escalations.

The report argues that these results show fewer, more targeted alerts can improve the identification of genuine risks. It also found that banks making the most progress are moving away from static, rules-based detection models toward approaches that examine behavioural patterns.

According to the findings, those institutions are also bringing fragmented data and workflows into more integrated systems. Another distinguishing feature is that they are linking surveillance spending to measurable reductions in manual review work.

For banks that have yet to make that shift, the challenge appears to be less about access to AI tools and more about whether internal systems and processes can support them. The research suggests that without cleaner data, integrated workflows and less reliance on outdated infrastructure, AI spending alone is unlikely to reshape surveillance operations.

That matters because communications surveillance has become a core function within bank compliance frameworks, particularly as firms manage a growing range of digital channels. High alert volumes, heavy review burdens and fragmented data can raise costs while making it harder to isolate real conduct and compliance risks.

Industry response

Shiran Weitzman, Co-Founder and Chief Executive Officer of Shield, said the main problem was no longer a lack of intent to invest.

"AI ambition is everywhere right now. What's missing is execution," said Shiran Weitzman, Co-Founder and Chief Executive Officer of Shield. "Investment is accelerating, but the fundamentals still matter. AI cannot deliver its full potential when surveillance is constrained by fragmented data, legacy infrastructure and alert-heavy operating models. Closing that gap is what will separate firms that simply deploy AI from those that fundamentally improve how risk is detected and understood."

The findings suggest banks are entering a new phase of surveillance technology adoption, where implementation and operational design may matter more than headline spending plans. For compliance leaders, the benchmark now appears to be whether investment produces clearer detection outcomes and reduces the manual burden on staff.

The report says surveillance is moving toward a model built around integrated data, context-driven detection, reduced operational noise and measurable outcomes. It found that institutions closing the gap between available technology and actual practice are better placed to lower costs while improving true risk detection.