AI uptake is surging ahead in the UK’s financial service industry as firms begin to look for more diverse use cases for the tech, but workers are calling for stronger oversight to avoid hidden risks, according to a new study from Smarsh.

The digital communications platform found that over a third (37%) of financial service employees regularly use public AI tools like ChatGPT or Microsoft 365 Copilot, despite 55% admitting they’ve had no formal training. 

Worryingly, 38% of employees reported that they aren’t sure whether their organisation is monitoring AI outputs, while 21% said they’re certain that their firm does not.

While other studies have suggested that many organisations are struggling to cope with burgeoning shadow AI, Smarsh’s research found that those working in financial services would prefer more guardrails rather than fewer.

The majority (70%) of employees agreed that they’d feel more confident using AI tools if all outputs were captured and monitored for transparency, privacy and compliance issues.

Added to that, a third (31%) believe their firms might lack the regulatory safeguards needed to manage these tools, and 29% would like more clarity around where potentially sensitive information is going when AI agents are used.

Employee concerns don’t seem to be getting much traction, however, with nearly half (43%) saying their organisations now use fully automated AI agents for tasks like customer communications, while 22% say these tools are involved in investment activities.

Recommended reading

• FCA Data Chief on the ‘Spinning Coin’ of AI in Financial Services
• FCA Warns Finance Firms About AI Fraud Risks
• 90% of Finance Functions to Deploy AI by 2026

With research from the Bank of England showing that 75% of UK financial firms now use AI to power everything from customer service and risk management to anti-money laundering and cybersecurity, Smarsh wants businesses to rethink their AI responsibilities before deploying more consumer-facing tools.

“Using public Al tools without controls is digital negligence,” said Paul Taylor, vice president of product at Smarsh. 

“You’re effectively feeding your crown jewels into a black box you don’t own, where the data can’t be deleted, and the logic can’t be explained. It’s reckless. 

“Private tools like Microsoft 365 Copilot and ChatGPT Enterprise are a step in the right direction. Still, if companies aren’t actively capturing and auditing usage, they’re not securing innovation- they’re sleepwalking into a compliance nightmare.”