Delivering Good Client Outcomes Through Technology: FCA Consumer Duty and AI (2026)
Serra Education | AI-for-Wealth insights | UK
The FCA Consumer Duty asks a hard question of every UK wealth firm: can you show that your clients are getting good outcomes, not just that you followed a process? Artificial intelligence sits on both sides of that question. Used without governance, it is a Consumer Duty liability waiting to be found in a supervisory review. Governed correctly, with validation, a data boundary, and an audit trail, it becomes the most efficient way a firm can evidence good outcomes across its whole book. This piece walks the four Consumer Duty outcomes and shows, for each, how a governed AI process improves the outcome while keeping a named human accountable.
This is process and governance guidance, not legal, investment, or compliance advice. Confirm every obligation against the primary source and your own counsel. For the wider governance frame, see the AI-for-Wealth hub for UK wealth firms.
What is the FCA Consumer Duty and where does it live in the Handbook?
The Consumer Duty is a set of FCA rules requiring firms to act to deliver good outcomes for retail customers. It sits in the Handbook at PRIN 2A, is underpinned by the Consumer Principle (Principle 12), and is explained in finalised non-Handbook guidance FG22/5 (July 2022). It raises the standard from treating customers fairly to demonstrably delivering good outcomes.
The Duty is built on three layers. A consumer principle (Principle 12: a firm must act to deliver good outcomes for retail customers). Cross-cutting rules: act in good faith, avoid foreseeable harm, and enable customers to pursue their financial objectives. And four outcomes that give the Duty its teeth: products and services, price and value, consumer understanding, and consumer support. The detail lives in PRIN 2A of the FCA Handbook (the retail customer outcome rules at PRIN 2A.3 to PRIN 2A.6), and the FCA's expectations are set out in finalised guidance FG22/5.
The word that matters for AI is demonstrably. The Duty does not only ask whether an outcome was good. It asks whether the firm can show the work: the monitoring, the evidence, the accountable owner. That evidentiary burden is exactly where a governed AI process earns its place, and where an ungoverned one becomes a finding.
Does the FCA have separate AI rules that override the Consumer Duty?
No. The FCA has taken a technology-neutral, no-new-rules approach to AI. It remains outcomes-based and technology-neutral, is not introducing new rules for AI, and expects firms to apply existing frameworks (the Consumer Duty, SM&CR, and operational resilience) to AI use. So the question is never whether the AI is compliant in the abstract, but whether the AI-assisted process still delivers good outcomes and keeps a human accountable.
The FCA's position, set out in its published AI Update and its wider AI approach and reflected in joint Bank of England and FCA work on AI in financial services, is that the existing regime already governs AI. There is no separate switch to flip. A firm cannot point to a model card or a vendor's assurance and call the matter closed. It must show that the outcome the client received meets PRIN 2A, that a named individual under SM&CR owns the outcome, and that the records under SYSC 9.1 exist to prove it.
For wealth firms this is clarifying, not frightening. You already know how to run a governed process under the Handbook. AI does not exempt you from that regime, and it does not demand a parallel one. It has to fit inside the conduct rules you already operate under. Serra's whole thesis follows from this: the blocker to using AI in a UK wealth firm is not the model, it is governance. Validation, a data boundary, and an audit trail are what turn AI from a Consumer Duty liability into an asset. For the fuller version of that argument, see why the AI gap in wealth is governance, not adoption.
Related reading: the FCA AI rules UK wealth advisers must apply in 2026.
How does AI help deliver the products and services outcome (PRIN 2A.3)?
The products and services outcome requires that products are designed to meet the needs, characteristics, and objectives of an identified target market and are distributed appropriately (PRIN 2A.3). Governed AI helps by processing target-market fit, suitability signals, and distribution data at a scale manual review cannot reach, flagging mismatches for a human to judge. The human still owns the design and distribution decision.
Under PRIN 2A.3, a manufacturer must identify a target market and ensure the product is consistent with the needs of that market, and a distributor must have distribution arrangements consistent with it. In practice this means continuously checking that the clients actually holding a product still sit inside its target market, and that clients outside it were not sold in.
A governed AI process reads the client base against the target-market definition and surfaces the exceptions: the client whose circumstances have drifted, the product being distributed into a segment it was never designed for, the cohort that never fit. It does the reading. It does not make the call. A human, accountable under SM&CR, reviews each flag and decides. What changes is coverage and speed: same headcount, a quarter of the hours, still inside your conduct rules, with every flag and decision written to the audit trail so the firm can evidence the review took place.
The governance conditions that make this safe are the same three every time. Validation: the target-market logic is tested against known cases before it runs on live clients. A data boundary: client data used for the check stays inside a controlled environment and is not exposed to an external model that would use it for training. An audit trail: every flag, every human decision, and every override is recorded so a supervisor can reconstruct what happened.
How does AI help deliver the price and value outcome (PRIN 2A.4)?
The price and value outcome requires firms to assess whether the price a customer pays is reasonable relative to the benefits received, and to identify products that do not provide fair value (PRIN 2A.4). Governed AI accelerates the fair-value assessment by aggregating cost, fee, and benefit data across the book and flagging thin-value cases, so the human starts from evidence.
PRIN 2A.4 requires a value assessment: a product provides fair value where the amount paid is reasonable relative to the benefits, and a firm must be able to show it. For a wealth firm this means comparing all-in cost (management fees, platform charges, transaction costs, adviser charges) against the service and benefit the client receives, across many client segments, and refreshing it regularly. Done by hand, it is a periodic, sampled, painful exercise. That is precisely why poor-value pockets go unnoticed between reviews.
A governed AI process turns the value assessment into something continuous. It assembles the cost stack per client, compares it against the benefit and against comparable offerings, and ranks where value looks thin. The fair-value judgement stays a human one, made by the accountable owner, because "reasonable relative to benefits" is a judgement the FCA holds a person responsible for. But the person now makes it from a complete, current evidence base instead of a spreadsheet sample. The audit trail then holds the assessment, the data behind it, and the decision, which is exactly the evidence FG22/5 expects a firm to produce on request.
How does AI help deliver the consumer understanding outcome (PRIN 2A.5)?
The consumer understanding outcome requires firms to communicate in a way that equips customers to make effective, timely, and properly informed decisions, and to test and adapt communications (PRIN 2A.5). Governed AI helps draft, tailor, and test communications for clarity at scale, while a human approves every client-facing message before it goes out. AI drafts; the human signs.
PRIN 2A.5 asks firms to support informed decision-making: communications must be clear, fair, and not misleading, and firms are expected to test whether communications are actually understood. This is where AI is most obviously useful and most obviously dangerous. Useful, because a language model can turn a dense suitability rationale or a fee disclosure into plain, layered explanation and adapt it to a client's circumstances. Dangerous, because an ungoverned model can also invent a figure, soften a risk, or stray from non-advice into an implied recommendation.
The governance answer is a firm approval gate, not a ban. A governed process uses AI to draft and to stress-test communications for readability, then routes every client-facing output through a named human who checks it against the facts and the perimeter before it is sent. Validation here means the drafting process is tested against known-correct source material so it does not fabricate. The data boundary keeps client specifics inside the controlled environment. The audit trail records the draft, the human edits, and the approval, so the firm can show a supervisor both that the communication was clear and that a person stood behind it.
This is the same discipline Serra applies to its own content: a piece is drafted with AI, verified against named primary sources, and approved by a human before anything ships. The tooling scales the drafting. It does not replace the accountable signature.
How does AI help deliver the consumer support outcome (PRIN 2A.6)?
The consumer support outcome requires firms to provide support that meets customers' needs and to ensure customers do not face unreasonable barriers, so that acting on an instruction, complaining, switching, or cancelling is not unreasonably harder than buying was (PRIN 2A.6). Governed AI helps triage, route, and speed up support and complaints handling, catching vulnerable-client signals early, while human judgement handles the substance.
PRIN 2A.6 sets an even-handedness standard for support: acting on an instruction, raising a concern, or switching away must not be made harder than buying in the first place. AI can materially improve support responsiveness by triaging inbound queries, surfacing the client's full context to the person handling them, and flagging the signals of a vulnerable client (a bereavement, confusion, financial distress) that a busy inbox can miss. Faster, better-informed support is a good outcome, and the Duty rewards it.
The line to hold is that support triage is not support substitution. A governed process lets AI classify, prioritise, and prepare, but a human owns any response that affects the client's money or rights, and certainly any complaint outcome. The vulnerable-client flags in particular must escalate to a person, because the Duty treats characteristics of vulnerability as requiring heightened, human care. The audit trail records the triage, the escalation, and the human handling, which is what lets a firm evidence, under FG22/5's monitoring expectations, that its support met client needs.
Who is accountable when AI is in the process (SM&CR)?
A person is, always. The Senior Managers and Certification Regime (SM&CR) requires a named senior manager accountable for each area of a firm's business, and AI does not create an exception. The FCA's no-new-rules approach means an AI-assisted outcome has the same accountable owner it would have had without the AI. The model is a tool inside that responsibility.
Under SM&CR, senior managers hold statements of responsibility and a duty of responsibility for the areas they run. When a process uses AI, the FCA's technology-neutral stance means accountability does not diffuse into the vendor or the model. The senior manager responsible for the outcome stays responsible. Conduct Rule 6 under the Consumer Duty reinforces this at the individual level: a person must act to deliver good outcomes for retail customers.
The practical consequence is that "the AI decided" is not a defence. Every governed AI process Serra helps design keeps a specific, named human in the accountable seat, with the authority to override the model and the responsibility to review its output on anything consequential. The audit trail exists in part to protect that person: it lets them show what the model produced, what they decided, and why. Governance does not brake the senior manager. It is the evidence that lets them stand behind the outcome.
What records does the FCA expect for AI-assisted decisions (SYSC 9.1)?
SYSC 9.1 requires firms to keep orderly records of their business and internal organisation, sufficient to enable the FCA to monitor compliance. For an AI-assisted process, that means recording the inputs, the model's output, the human's decision, and any override, in a form a supervisor can reconstruct. The audit trail is not optional housekeeping; it is the compliance artefact.
SYSC 9.1.1R requires a firm to arrange for orderly records to be kept of its business and internal organisation, sufficient to enable the FCA to monitor the firm's compliance with the requirements of the regulatory system and, in particular, to ascertain that the firm has complied with all obligations with respect to clients. When AI sits inside a decision, adequate records extend to the AI-assisted step. In practice a firm should be able to show, for a given outcome: what data went in, what the model returned, who reviewed it, what they decided, and what changed as a result. This is the same recordkeeping logic the Consumer Duty's monitoring obligation depends on, and the same logic FG22/5 assumes when it expects firms to evidence good outcomes.
This is the crux of Serra's thesis and the reason governance flips the equation. An ungoverned AI process is a Consumer Duty liability precisely because it produces outcomes it cannot evidence: no reconstructable record, no accountable reviewer, no boundary on the data. A governed one produces the record as a by-product of the process. The audit trail that satisfies SYSC 9.1 is the same trail that lets the firm demonstrate good outcomes under PRIN 2A. Build the governance once, and it serves both.
How does Serra help a wealth firm make AI a Consumer Duty asset?
Serra Education provides process and governance consulting for AI adoption in wealth firms. It is vendor-neutral: Serra does not resell an AI stack. The work is designing the validation, the data boundary, and the audit trail that let a firm use AI inside the Consumer Duty and SM&CR, so the same team covers far more of the outcomes work in far fewer hours.
Serra's engagement is deliberately narrow and deliberately not a technology sale. Serra does not sell you a model, take a margin on a platform, or push a particular vendor. It maps your existing Consumer Duty obligations under PRIN 2A onto the points in your process where AI can safely do the reading and the drafting, then builds the three governance conditions that keep it compliant: validation of the AI logic against known cases, a data boundary that keeps client data out of external training, and an audit trail that satisfies both SYSC 9.1 and your Consumer Duty monitoring in one artefact.
The outcome we aim for is the line that runs through everything Serra does: same headcount, a quarter of the hours, still inside your conduct rules. The four Consumer Duty outcomes stop being a periodic, sampled, manual burden and become a continuously monitored, evidenced, human-owned process. That is what makes AI a Consumer Duty asset instead of a liability. And to be explicit about the perimeter: this is governance and process work. Serra Education does not provide Serra Wealth investment advice through it, and nothing here is a recommendation on any product or portfolio. To see how this fits the broader capability, start at the AI-for-Wealth hub.
Frequently asked questions
Does using AI breach the FCA Consumer Duty?
No, not in itself. The FCA's technology-neutral, no-new-rules approach means AI is governed by existing frameworks, including the Consumer Duty (PRIN 2A) and SM&CR. AI breaches the Duty only when it produces outcomes a firm cannot evidence or that no accountable human owns. Governed correctly, it helps deliver and evidence good outcomes.
Do I need FCA approval to use AI in my wealth firm?
There is no separate AI authorisation. The FCA has said it is not introducing new rules for AI and expects firms to apply the existing regime. You do not seek AI-specific approval; you ensure your AI-assisted processes meet your existing obligations under PRIN 2A, SM&CR, and SYSC. Confirm your firm's specific permissions with your compliance function and counsel.
Can AI make an advice or suitability decision for a client?
Under the FCA regime the accountable decision stays with a human. SM&CR requires a named individual to own the outcome, and Conduct Rule 6 requires that person to act to deliver good outcomes for retail customers. A governed process uses AI to prepare, flag, and draft, and keeps the consequential suitability or advice decision with the accountable person.
What is the single most important control for AI under the Consumer Duty?
The audit trail. SYSC 9.1 requires records adequate for the FCA to monitor compliance, and the Consumer Duty requires firms to evidence good outcomes. A reconstructable record of the inputs, the model output, the human decision, and any override satisfies both at once, and it is what turns an AI process from an unevidenced liability into a demonstrable asset.
Is this Serra Wealth investment advice?
No. This article and Serra Education's AI-governance consulting are process and governance guidance only. Serra is vendor-neutral, does not resell an AI stack, and does not provide Serra Wealth investment advice through this work. Nothing here is a recommendation on any product, portfolio, or regulatory position.
AI disclosure. This article was drafted with AI assistance and reviewed by a human before publication, the same governed process it describes.
Not advice. This is process and governance guidance, not legal, investment, or compliance advice. Confirm every obligation against the primary source (the FCA Handbook PRIN 2A and SYSC 9.1, FCA finalised guidance FG22/5, and the FCA's published AI approach) and your own counsel.
Primary sources referenced: FCA Handbook PRIN 2A (Consumer Duty), including Principle 12 and PRIN 2A.3 to 2A.6; FCA Consumer Duty Individual Conduct Rule 6 (COCON 2.4); FCA finalised guidance FG22/5 (Consumer Duty, July 2022); FCA Handbook SYSC 9.1 (record-keeping); the Senior Managers and Certification Regime (SM&CR); the FCA's published approach to AI (AI Update / "AI and the FCA: our approach").