A warning from the intermediary sector is that AI may take substantial administration “off your desk”, but not the responsibility when things go wrong. That message aligns closely with the FCA’s latest Consumer Duty observations: stronger firms support fair-value conclusions with meaningful evidence, assess outcomes across different customer groups and test whether customers understand the price and commitment involved.
For mortgage and financial-advice firms adopting generative or agentic AI, the governance question is not simply whether a tool saves time. It is whether the firm can evidence how outputs are checked, when human judgement is required, how vulnerable customers are protected and whether technology-driven processes create consistently good outcomes.
- Build controls around: data quality, source records, human approval and exception handling.
- Evidence outcomes: use complaints, customer feedback, arrears data and case-file review to test impact.
- Avoid over-reliance: an AI-generated summary or recommendation rationale should be reviewed, not accepted automatically.