At a Glance
- What this covers: What FCA Consumer Duty actually requires from WealthTech AI, why most platforms fail the compliance review, how governance-first SDLC resolves the gap, and three self-assessment questions to run before your next Consumer Duty review cycle.
- Key finding: FCA Consumer Duty changed what regulated AI must do – not just perform well, but prove it. Most WealthTech compliance teams discovered the gap during their first Consumer Duty review cycle. The AI was performing. The governance infrastructure to demonstrate compliance was not there.
- Business impact: The four Consumer Duty outcome requirements cannot be demonstrated retrospectively for an AI system not built to produce the evidence. Consumer Duty compliance is an engineering requirement – not a documentation exercise that can be completed after examination is announced.
- What you will learn: The four Consumer Duty outcome requirements for WealthTech AI, the three engineering gaps that cause most review failures, and the five SDLC deliverables that make compliance a byproduct of development.
FCA Consumer Duty changed what regulated AI must do – not just perform well, but prove it. Every AI recommendation a WealthTech platform makes must be accompanied by suitability evidence, fair value documentation, and an explainable outcome that can be produced on demand. Not reconstructed. Produced. At inference time.
Most wealthtech compliance teams discovered this gap during their first Consumer Duty review cycle. The AI system was performing. The governance infrastructure required to demonstrate compliance was not there – because it was never built. Consumer Duty compliance is an engineering requirement. It cannot be completed after the examination is announced.
I. What FCA Consumer Duty actually requires from WealthTech AI
The Duty came into force on 31 July 2023 for new and existing products open for sale, and on 31 July 2024 for closed products and services. If you run legacy portfolios or a closed book, they are in scope.
Everything sits under Principle 12: firms must act to deliver good outcomes for retail customers. Beneath it are three cross-cutting rules that act in good faith, avoid foreseeable harm, and enable customers to pursue their financial objectives.
The four outcomes below are how that principle is tested in practice.
- Products and services outcome: AI-powered products must be designed to deliver good outcomes for retail customers – with documented evidence the design process considered customer needs, not just product performance.
- Price and value outcome: Fair value must be demonstrable at the point of recommendation, with a documented methodology rather than a justification written after the fact. For AI-driven pricing and product selection, this means the value assessment has to be a recorded input to the decision, not a periodic review sitting alongside it.
- Consumer understanding outcome: AI outputs must be explainable to retail customers in plain terms. A recommendation produced by a black-box model that cannot explain its own reasoning fails this requirement regardless of accuracy.
- Consumer support outcome: Customers must be able to act on, question, switch or complain about an AI-driven outcome without facing unreasonable barriers. Where AI handles or triages support, the firm has to show that it does not make the exit harder than the entry. Suitability evidence supports this outcome, but it is not the outcome itself.
None of these four outcomes can be demonstrated retrospectively for an AI system not built to produce the evidence.
II. Why most WealthTech AI fails the Consumer Duty compliance review
No AI suitability evidence at inference
Consumer duty compliance requires suitability evidence per recommendation – the specific record of why this AI output was appropriate for this customer profile, at this time, given these stated objectives. Most WealthTech AI systems produce a recommendation. None of the underlying inference records is captured in a queryable format. When the FCA requests AI suitability evidence for a specific customer, the engineering team has nothing to provide.
No explainability at the model layer
Consumer understanding requires AI recommendations to be explainable in plain terms. For AI financial advice systems, this means the model must produce a human-readable rationale for every recommendation – not approximated by a post-hoc tool, but generated as a byproduct of the inference process. Models not designed for explainability cannot produce this without fundamental re-architecture.
No fair value documentation at inference time
Price and value outcome requires documented evidence that the AI-powered product delivers fair value. For algorithmic pricing and product recommendation systems, this means capturing – at every inference – the value assessment methodology applied and the customer profile it was assessed against. Wealth management compliance at this level requires structured logging built into the inference pipeline from sprint one.
III. What Consumer Duty compliance looks like when built correctly
A UK-based WealthTech startup had AI embedded across their platform delivering investment recommendations, client onboarding, and portfolio management. When their first Consumer Duty review cycle began, the compliance team could not produce suitability evidence for specific past recommendations – the inference record had never been captured in a queryable format. Explainability required a separate post-hoc tool that had not been part of the original model design.
Systango rebuilt the governance infrastructure as engineering deliverables:
- explainability architecture designed before the model retrained
- suitability logging built into the inference pipeline
- fair value documentation encoded as governance events at every pricing inference, and
- independent validation structured before the next deployment cycle.
Outcome: production AI co-pilot live across four knowledge domains with full suitability evidence at inference, five core user journeys delivered with Consumer Duty documentation built in, governance layer active from day one of each new deployment.
Lesson from this engagement: the compliance team stopped being asked to reconstruct evidence that the engineering team had never captured. The moment suitability logging became part of the definition of done, the Consumer Duty review became a reporting exercise, not an investigation.
IV. Three questions to ask before your next FCA Consumer Duty review cycle
If any answer is no or uncertain, the gap is an engineering problem. It is buildable – but only before the model is deployed, not after the review cycle begins.
Key Takeaways
- No AI rulebook is coming. The FCA has confirmed it will govern AI through Principle 12, the Consumer Duty and SM&CR, which means the definition of adequate evidence is your firm’s to make and defend.
- Accountability is personal, not departmental. Under SM&CR it attaches to the senior manager for the business area, and it cannot be transferred to a model or a vendor.
- The annual board report is where the gap surfaces. If the AI cannot produce outcome data, the board cannot make a claim about it, and silence on your highest-volume decision system is itself a finding.
- Two of the three common failures are fixable on a live system. Logging and outcomes monitoring can be added. Model-layer explainability is architectural, so plan for it before the model trains, not after.
- The practical test is whether governance is inside the definition of done. When it is, Consumer Duty evidence is a byproduct of shipping. When it is not, it is a reconstruction project with a deadline attached
Systango’s AI-native SDLC embeds FCA Consumer Duty compliance into every WealthTech AI delivery: explainability architecture designed before the model trains, AI suitability logging built into the inference pipeline, fair value documentation encoded as governance events, and independent validation structured before the first sprint. As a publicly listed, ISO 27001 certified engineering company with active Consumer Duty delivery experience across WealthTech, explore our AI-native SDLC and AI Readiness Assessment to understand how we approach your specific compliance challenge.
