There is a wide gap between what banks say they want from surveillance technology and what they have actually managed to deploy.
According to 1LoD’s 2026 Surveillance Benchmarking Survey and Report, around 89% of banks want AI-enhanced trade surveillance, yet only 11% have it running. Some 78% want generative AI assistants for their analysts, but just 7% have one in place, said Wordwatch.
No bank surveyed reported having AI fully embedded as part of its target operating model in either trade or holistic surveillance, with e-comms adoption at 8% and voice at 9%.
Budget and regulatory caution are the usual excuses, but the survey undermines both: 78% of banks say they have sufficient budget to run surveillance effectively, and 67% say sanction risk no longer limits their move towards risk-based approaches, up from 59% two years ago.
The real constraint sits upstream, in the data itself. Asked which issue most hinders their surveillance systems, 71% of responses pointed to fragmented data across silos, a lack of standardised formats or identifiers, and inconsistent data quality.
Limited access to quality data is rated a high or medium challenge by 78% of banks, and legacy surveillance systems rate similarly for four in five institutions. Governance, by contrast, has largely settled, with only 7% citing unclear regulatory expectations as a high challenge.
Layering more analytics onto poor data does not fix the problem, it just multiplies the noise. False positives are flagged as a meaningful drag by 93% of banks. The survey traces this back to fragmented capture, inconsistent trading data and ageing vendor platforms sitting upstream of alert generation.
Holistic surveillance shows the clearest strain: 48% of banks have not linked any surveillance controls, and none combine trade and communications data before the alert stage, because there are no reliable common identifiers to join them on.
A second, harder problem follows. Around 37% of banks cite difficulty validating NLP and large language model based systems as a top model risk challenge, alongside limited internal validation expertise. Richard Littlechild, head of secondary market oversight at the Financial Conduct Authority, writes in the report’s foreword that fundamentals matter more as the environment evolves, describing technology as “an enhancer, not a replacement.”
Before AI can cut review burden, records need to be complete, held in original format, traceable through chain of custody, and reconciled against source. That work typically falls outside the surveillance budget entirely, which is why so many AI initiatives remain stuck at proof of concept.
Read the full Wordwatch post here.
Copyright © 2026 RegTech Analyst
Copyright © 2026 RegTech Analyst
