FSCA's Treating Customers Fairly (TCF) framework is the conduct lens through which SA financial services regulation evaluates customer outcomes. AI-driven customer interactions are evaluated through the same framework as human-driven ones.
This article walks through what each of the six TCF outcomes requires when AI is part of how the firm delivers products, services, and support to customers.
Outcome 1: TCF central to culture
Customers can be confident they are dealing with firms where TCF is central to corporate culture. Culture extends to AI deployment decisions and AI operating discipline.
AI implications:
● AI use cases evaluated against customer outcomes, not only commercial outcomes
● AI deployment decisions that prioritise commercial gain over customer fairness affect the cultural outcome at a structural level
● Board and senior management engagement with AI as a TCF consideration, not only as a technology or risk consideration
● Workforce understanding that AI tools they use carry TCF obligations the firm is accountable for
Outcome 2: Products and services designed for identified customer groups
Products and services marketed and sold are designed to meet the needs of identified customer groups and are targeted accordingly. AI affects this outcome through product recommendation, customer segmentation, AI-assisted product design, and targeting decisions.
AI implications:
● AI-driven product recommendation must surface products genuinely fitted to identified customer needs, not products optimised for the firm's commercial outcomes at the expense of fit
● Customer segmentation driving differential offering needs review for fairness and alignment with customer needs
● AI-assisted product design must reflect customer needs rather than only firm metrics
● Target market definitions need to account for AI-driven targeting decisions, including who AI effectively excludes
● Vulnerable customer identification through AI must be carefully designed, AI may both improve identification and introduce its own biases
Outcome 3: Clear information at point of sale and beyond
Customers are provided with clear information and are kept appropriately informed before, during, and after the point of sale. AI affects this outcome through AI-generated customer communications, chatbots, AI-assisted advisers, and AI-driven personalisation.
AI implications:
● AI-generated customer communications must be designed to support understanding, not just technical compliance
● Chatbots and AI agents serving customers need to produce content customers can act on, with appropriate escalation when AI cannot deliver the needed clarity
● AI-driven personalisation that adjusts information to different customers must be reviewed for whether it supports understanding or fragments it inappropriately
● Vulnerable customers may need specific accommodation, AI communications optimised for the average customer may fail vulnerable customers in ways that breach TCF
● Disclosures required by FAIS and other sector frameworks must be delivered effectively, not just technically, by AI-driven channels
Outcome 4: Suitable advice
Where customers receive advice, the advice is suitable, taking account of their circumstances. AI-driven advice or AI-augmented advice must satisfy suitability.
AI implications:
● AI-driven product recommendation that crosses into advice territory must satisfy FAIS advice requirements
● AI-augmented advisers must operate within FAIS suitability framework, with AI inputs neither substituting for nor undermining the advice process
● Robo-advisory and similar AI-driven advice services have specific regulatory considerations beyond general TCF
● Documentation of advice suitability must extend to AI-driven components
● Vulnerable customer protections in advice are reinforced, not weakened, by AI involvement
Outcome 5: Products and services performing as led to expect
Products perform as firms have led customers to expect, and the service is of an acceptable standard. AI affects this outcome through AI-driven service delivery.
AI implications:
● AI-driven service delivery, chatbots, AI agents, AI-assisted human agents, automated triage, must meet expected service standards
● Failures in AI delivery, hallucination, incorrect information, inability to help, are TCF failures regardless of cost efficiency
● Service monitoring covering AI-driven service must use the right metrics, not just operational AI metrics but service quality, customer outcome, complaint patterns
● Vulnerable customer service through AI channels needs specific attention
● AI service that worked at deployment may degrade through model updates; ongoing monitoring is part of TCF discipline
Outcome 6: No unreasonable post-sale barriers
Customers do not face unreasonable post-sale barriers to change product, switch provider, submit a claim, or make a complaint. AI in retention, claims handling, complaint management must not create unreasonable barriers.
AI implications:
● AI in claims handling must not create unreasonable obstacles to legitimate claims through automated denial, excessive verification, or opaque processing
● AI in retention must not create unreasonable obstacles to switching, friction added by AI is still friction
● Complaint handling involving AI must support effective complaint resolution, not deflect or delay through AI-mediated bottlenecks
● Escalation from AI to human channels for customers seeking to leave, claim, or complain must be accessible and effective
Vulnerable customers across the TCF outcomes
Vulnerable customer considerations extend across all six TCF outcomes when AI is involved. AI may help identify and serve vulnerable customers; AI may also systematically fail vulnerable customers in ways that aggregate metrics obscure.
● AI identification of vulnerability, useful for consistency, requires validation for bias
● AI interactions with identified vulnerable customers, must adjust, not just continue with vulnerability flagged
● Escalation to specialist support, reliable when AI identifies vulnerability need
● Outcome monitoring, must specifically cover vulnerable customer experience, not aggregate that obscures
● Testing methodology, vulnerable-customer-relevant test cases as part of standard AI evaluation
Documentation and evidence
TCF operates on evidence. AI-related TCF obligations require specific documentation:
● AI inventory mapped to customer journeys and to TCF outcomes
● Outcome testing methodology for AI-driven elements
● Monitoring metrics covering AI-driven customer outcomes
● Management information for senior management oversight of AI TCF performance
● Complaint analysis covering AI-related customer issues
● Improvement actions documented and tracked
The shift to make
Stop treating TCF as a customer journey exercise that AI happens to be part of.
Start treating it as the customer outcome framework AI must specifically deliver against, with AI-specific operational discipline, AI-specific testing, AI-specific outcome monitoring, and AI-specific senior management oversight.
Firms operating this way pass FSCA TCF engagement on AI as a matter of course. Firms operating AI as a separate workstream from TCF discover, through supervisory dialogue or complaint patterns, that the integration gap is exactly what the FSCA examines.
