Skip to main content
Curvestone AI
How to

Financial promotion approval: who signs off what the AI passes

Updated
  1. What the rules ask of a lender's promotions

    A lender's promotions sit under CONC 3 for consumer credit and MCOB 3A for mortgages. Under CONC 3.3.1R, each promotion must be accurate, must be "balanced", and must not "disguise, omit, diminish or obscure important information, statements or warnings". For an MCD regulated mortgage contract, MCOB 3A.2.4R also rules out wording that "may create false expectations for a consumer regarding the availability or the cost of a credit".

    None of these rules says how to check a promotion or who signs it off. That design is yours. What they ask is that you can show the steps you took.

    A common point of confusion is the approver gateway. The FCA says a firm "can only approve a financial promotion if the FCA has granted the firm permission to do so", or an exemption applies. Its page lists exceptions, including "approvals of an authorised person's own financial promotions for the purpose of communication by an unauthorised person". A lender that approves its own material for brokers to use does not usually need that permission, but the content rules still apply in full.

  2. Where the AI check sits in the workflow

    A practical approval workflow for a lender has six steps:

    1. Brief. Marketing records the product, the audience, the channel and the claims it wants to make. The brief tells the check which rules apply: CONC, MCOB 3A or both.
    2. Draft. Marketing produces the copy, image or video in its usual tool.
    3. AI check. The draft goes to the check before anyone in compliance opens it. The check reads the copy and the visual content against the rules and your house standard, and returns each issue with the rule it relates to.
    4. Marketing fixes. Marketing clears the routine issues (a risk warning in small print, a benefit with no balancing risk, a claim the brief cannot support) and resubmits.
    5. Compliance review and sign-off. A named approver reviews the promotion, the AI result and any issue marketing disputes. The approver approves, rejects or sends it back.
    6. After it goes live. The promotion is checked again when the product, the rate or the rules change, and on the schedule your risk appetite sets.

    If the AI check runs after compliance has approved, it adds cost and little else. Run before compliance, it removes the back-and-forth that slows most approval queues.

  3. Who signs off what the AI passes

    The same person who signed off before AI: a named approver with the knowledge and the authority to reject. What changes is what that person looks at.

    The FCA's October 2026 paper found "Excessive reliance on human-in-the-loop controls" and warned that "Human review may mask weaknesses in the model". In one case study, reviewers "increasingly accepted AI-generated recommendations without sufficient scrutiny".

    Three design choices reduce that risk:

    • Give the approver the evidence, not only the verdict. A pass with no reasons invites a rubber stamp. A result that shows each check and the rule behind it gives the approver something to test.
    • Set clear escalation rules. Decide which issues an approver can clear alone and which go to a senior compliance owner, for example new product claims, promotions to vulnerable groups or social media content.
    • Test the approvers as well as the tool. The FCA paper describes a firm where reviewers assessed cases "without being shown the AI's proposed outcome", so it could compare human and AI judgements. A periodic blind sample does the same for promotions.

    Accountability for the AI tool itself sits with a senior manager. The paper describes good practice where firms "mapped responsibilities for AI across senior management functions". In one case study, "Ultimate accountability sat with the SMF responsible for the business area the tool would operate in."

  4. The record that shows the control works

    The FCA paper describes one small firm that recorded "whether they approved the outputs, overrode them or escalated issues", which created "an audit trail showing how oversight operated in practice". For promotions, keep for each item:

    • the version of the promotion that was checked and the version approved;
    • the AI result, with each issue and the rule it cited;
    • the approver's name, the date and the decision;
    • the reason for each override, where the approver cleared an issue the AI flagged.

    This record answers the MCOB 3A.2.4R question of whether you can show reasonable steps. It also gives you management information: which rules fail most often, which channels produce the most issues, and how often approvers overrule the check.

    Curvestone checks financial promotions, including documents, social posts, images and video, against FCA rules and your firm's own standards, and returns a graded result with the rule cited for each issue, so the approver reviews evidence rather than a verdict. For the network view of the same question, see how AR oversight covers financial promotions, and for the wider FCA position, what the FCA's AI good practice paper means for lenders.

Questions

Frequently asked questions

Can an AI tool approve a financial promotion on its own?
The rules do not mention AI, but the firm stays responsible for the promotion and for any AI it uses. The FCA's October 2026 paper says firms are responsible for AI systems they procure and deploy. In practice, lenders keep a named, qualified approver who reviews the AI result and makes the decision.
Does a lender need FCA permission to approve its own promotions for brokers?
Usually not. The FCA lists approvals of an authorised person's own financial promotions, for communication by an unauthorised person, as an exception to the approver permission requirement. The promotion must still meet the content rules in CONC 3 or MCOB 3A, so the lender's own check and sign-off still apply.
What should a lender record when an approver overrides the AI check?
Record the issue the AI flagged, the rule it cited, the approver's name, the date and the reason for clearing it. Over time, these records show whether approvers are challenging the tool or deferring to it, and which rules your marketing team finds hardest to meet.
Sources
  1. 01FCA: AI in financial services: good practice and areas for improvement (19 October 2026)
  2. 02FCA Handbook: CONC 3.3 The clear fair and not misleading rule and general requirements
  3. 03FCA Handbook: MCOB 3A.2 The fair, clear and not misleading rules
  4. 04FCA: Approving financial promotions
Related reading
Dawid Kotur
Written by

Dawid Kotur

CEO and co-founder, Curvestone

Dawid co-founded Curvestone in 2024 after a decade working at the intersection of financial services and applied machine learning. He writes about the strategic direction of regulated-industry AI, the FCA's evolving approach to model risk, and the operational changes UK lenders are making in response to Consumer Duty. He sits on the FCA Smart Data Accelerator advisory cohort.

LinkedIn

Compliance that thinksahead. Automatically.

Join mortgage networks, lenders, and legal firms using Curvestone to review cases at scale.