FCA Consumer Duty and AI: What UK Financial Services Firms Must Do Now

Consumer Duty as an AI governance framework

The FCA's Consumer Duty, fully in force from July 2023, is outcome-focused rather than rules-based. It requires firms to deliver good outcomes for retail customers across four outcome areas: products and services, price and value, consumer understanding, and consumer support. This outcome focus makes Consumer Duty a powerful AI governance instrument, because it asks not whether the firm's AI systems are technically compliant with specific rules but whether they produce good outcomes for consumers.

This is a materially higher bar than traditional compliance. A firm can have an AI system that follows all the technical rules — it is not prohibited, it meets technical documentation requirements, it has been validated — but still fails Consumer Duty if it systematically produces outcomes that disadvantage consumers. The FCA has made clear that outcome-based assessment is the right approach: it is the consumer's experience that matters, not the firm's documentation.

The price and value outcome and algorithmic pricing

Consumer Duty's price and value outcome requires firms to offer products and services that represent fair value — the price charged must be reasonable given the benefits provided. For firms using algorithmic pricing, this creates a specific obligation to monitor whether pricing algorithms produce fair value for all customer segments. Loyalty penalties — where long-standing customers pay more than new customers for identical products — are an explicit Consumer Duty concern that the FCA has already signalled will receive supervisory attention.

More broadly, algorithmic pricing that systematically charges higher prices to customers based on their vulnerability, their limited alternatives, or their lower propensity to switch will fail the price and value outcome. The fact that the pricing is set by an algorithm rather than a human pricing decision does not reduce the firm's responsibility for the outcome — it increases the governance obligation to monitor what the algorithm is doing.