AI compliance layer could become essential for motor finance, says iVendi

Motor finance lenders and dealers may need to introduce a new layer of AI-powered compliance oversight as the use of agentic AI grows, if recommendations made to the Financial Conduct Authority (FCA) are adopted, according to iVendi.

The warning follows the publication of the Mills Review, which explores how artificial intelligence could reshape UK retail financial services. Among its recommendations are proposals to establish the foundations for agentic finance and for the FCA to develop its own AI-enabled supervisory model.

“The FCA Review could have significant implications for anyone involved in retail motor finance who is using, or plans to adopt, agentic AI,” he said.

“Our interpretation is that a second level of AI, ideally using completely separate technology, will be required for oversight of AI agents which are being used for any regulated activity.”

Under this approach, an independent AI compliance system would monitor customer interactions in real time, triggering a “kill switch” if it detected potential compliance concerns. Conversations would then be transferred immediately to a human adviser.

“If this oversight AI flags up an issue in real time, there would be a kill switch which ends the conversation immediately and passes the consumer to a human expert,” Tew explained. “It would also independently score every interaction and alert a senior manager where issues appear to have arisen.”

He added that organisations would also need to retain comprehensive records of every interaction handled by AI agents, mirroring the way many firms already record customer calls involving regulated activities.

“In exactly the same way as you might record every human call that takes place around regulated activity, you will ensure there are recordings and transcripts of everything the AI agent does. It’s evidencing that may be needed in the event of any complaint because ultimately, as now, overall responsibility lies with senior management.”

Tew believes this approach aligns closely with the FCA’s longer-term vision for AI supervision. The Mills Review recommends that the regulator eventually adopts its own “agentic supervisory model”, using AI-to-AI workflows to monitor firms continuously, assess compliance data and identify consumer harm or systemic risks more quickly than traditional reporting methods.

However, he noted that effective AI oversight will depend on consistent interpretation of the FCA’s Consumer Credit Sourcebook (CONC).

“AI agents base their dialogue on information from APIs provided by lenders and other parties, but their readings of the CONC Sourcebook do vary, and where there is room for interpretation, there is more likely to be hallucinations. This is an area that will need careful monitoring and management.”

Despite these challenges, Tew welcomed the direction of travel outlined in the review.

“Despite this, we think the Mills recommendations are welcome. There needs to be guidance on how agentic AI is used within motor finance and other areas of credit, and its general thinking is sound.”

He added that iVendi is already working with a range of lenders and dealers to support the development of AI agents using its API infrastructure.

“We’re able to provide data covering a wide range of areas, including finance quotes, eligibility checks and more, and can do so for a wide range of lenders, placing us in an almost unique position. It is an expanding area of our business.”