Short answer: the Duty already governs AI in retail claims
Yes. The Consumer Duty is in force for open retail products and services, and it applies to claims handling for in-scope insurance. Adding AI does not create a parallel rulebook. It raises the bar on how you evidence good outcomes, fair value, clear communications and usable support when machines sit in the journey.[1][2][4]
Last reviewed: 28 October 2026. Not legal advice (see the note at the end).
The FCA has been explicit that firms deploying AI-supported tools remain responsible for consumer outcomes, and that firms should read the Duty guidance (FG22/5) together with the vulnerable customer guidance (FG21/1).[4] Claims still have to be handled promptly and fairly under ICOBS 8.[7] Separately, a solely automated decline with a legal or similarly significant effect engages UK GDPR Article 22.[8][10] This post stays on the Duty. For how processing, handling and adjudication differ in practice, see AI claims processing explained. For EU high-risk classification, see Does the EU AI Act apply to UK insurers?.
What the Duty requires when AI sits in a claims journey
Treat every AI step as part of the customer journey the Duty already governs. The Consumer Principle, the three cross-cutting rules and the four outcomes all still apply; AI changes how you design, monitor and evidence them, not whether they exist.[1][2]
PS22/9 set a Consumer Principle that firms act to deliver good outcomes for retail customers, plus cross-cutting rules to act in good faith, avoid causing foreseeable harm, and enable customers to pursue their financial objectives. The four outcomes cover products and services, price and value, consumer understanding, and consumer support.[1]
- Products and services: claims processes must meet the needs of the target market, including how AI triages, requests documents or routes cases.
- Price and value: a claim that is slower, harder or less likely to pay because of AI design erodes the value the customer paid for.
- Consumer understanding: customers need to know what the AI did, what it could not do, and what to do next, in language they can use.
- Consumer support: support must meet needs when customers claim, including a usable route to a person when the automated path fails.
FG22/5 also expects claims not to be harder to pursue than cover was to buy. An AI journey that adds friction, dead ends or unexplained refusals sits poorly against that expectation and against ICOBS 8's duty to handle claims promptly and fairly and not to reject them unreasonably.[2][7]
Where AI helps outcomes, and where it creates Duty risk
Used well, AI shortens waits, spots missing evidence earlier and frees handlers for complex cases. Used poorly, it embeds bias, hides the reason for a decision, or locks customers into a channel they cannot leave. The Duty test is the outcome for the customer, not the sophistication of the model.[4][5]
| Duty outcome | AI that can help | Duty risk if unmanaged |
|---|---|---|
| Products and services | Faster triage to the right queue; earlier detection of incomplete files. | Wrong routing that delays urgent or complex claims; automation that excludes people who need a different path. |
| Price and value | Shorter settlement times that deliver the cover customers paid for. | Processes that deter claims, or settlement patterns that leave some groups worse off without objective justification. |
| Consumer understanding | Clearer status updates and plain-English next steps, with human review of wording. | Technical model talk instead of what the decision means; overstated accuracy or vague chatbot answers. |
| Consumer support | 24/7 progress updates and faster document collection for straightforward cases. | No easy exit to a person; journeys that repeat questions or stall after an automated refusal. |
Framed from FG22/5 outcomes and the FCA's October 2026 AI good-practice themes on Consumer Duty and vulnerable customers.[2][4]
The FCA's AI Update already warned that AI which embeds or amplifies bias, leading to worse outcomes for some groups unless differences can be justified objectively, may not be acting in good faith under FG22/5.[5] That warning lands hard in claims: decline rates, settlement values and complaint root causes need to be watched by customer group, not only in aggregate.
Vulnerable customers in automated claims journeys
Vulnerability is where many AI claims designs fail the Duty. FG21/1 expects firms to understand vulnerability needs, equip staff to respond, design flexible support and communications, and monitor whether vulnerable customers get outcomes as good as others.[3]
The FCA's AI good-practice publication ties that guidance directly to AI deployments. Good practice includes spotting vulnerability indicators and escalating to a person; areas for improvement include excluding vulnerable customers from AI journeys by default without an alternative path, and inflexible automated interfaces that make support harder to get.[4]
In the insurance multi-firm review of outcomes monitoring, the FCA also found that firms did not consistently monitor outcomes for distinct groups, including customers with characteristics of vulnerability. In one good-practice example, a firm found vulnerable customers had claims declined more often, investigated, and changed processes and training.[6] If your AI journey cannot show that comparison, you are not yet evidencing the Duty for claims.
What evidence the FCA expects boards to see
Boards need a causal chain from defined outcomes to metrics, investigation and remediation, not a stack of green process ticks. For claims AI, that means live outcome monitoring, meaningful human oversight of higher-impact decisions, and explanations customers and complaints handlers can use.[2][4][6]
The June 2024 insurance multi-firm review found many firms still monitored process completion rather than outcomes, and saw little monitoring of whether claim settlements were delivering good outcomes overall. Good practice showed clearly defined outcomes, a suite of metrics, identification of poor outcomes, action, and then evaluation of whether the action worked.[6]
For AI specifically, firms that use it safely and responsibly, in the FCA's words, define the consumer need before building, test outcomes across consumer characteristics, manage bias and accuracy with proportionate controls, explain decisions, tell customers how AI was used and its limits, and monitor live outcomes so issues are spotted quickly.[4] Human oversight must be real: reviewers need authority, information and the ability to challenge, not a rubber stamp that creates automation bias.[4]
Where a decision is solely automated and has a legal or similarly significant effect, Article 22 of the UK GDPR also requires an applicable exception and safeguards, including information for the individual and a simple way to obtain human intervention and contest the decision.[8][9][10] Token review after the fact does not make the decision "not solely automated".
An outcomes checklist for claims AI
Start from the customer outcome you owe under the Duty, then ask whether the AI design, the human fallback and the board MI can show that outcome is delivered. Six checks cover the ground most claims AI programmes miss.
Name the good and poor outcomes
For each AI step in the claims journey, write the customer outcome you intend and the foreseeable harms you must avoid, including for vulnerable groups.
Keep a usable human path
Customers must be able to reach a person without repeating their story. Declines and other higher-impact decisions need reviewers with authority to change the outcome.
Explain in customer language
Say that AI was used, what it could and could not do, why the outcome followed, and what the customer can do next. Test understanding, do not assume it.
Monitor outcomes by customer group
Track decline rates, settlement patterns, journey times, withdrawals and complaint root causes for vulnerable and other distinct groups, not only firm-wide averages.
Show the causal chain to the board
MI should move from metric to investigation to action to re-test. Process-completion scores alone will not satisfy PRIN 2A.9 or FG22/5 Chapter 11.
Align Article 22 and ICOBS 8
If any decision is solely automated with significant effect, document the Article 22 basis and safeguards. Keep claims prompt, fair and not unreasonably rejected under ICOBS 8.
If you want the operational shape of an AI claims stack that keeps a person on every decline, see Claim It Quick and the healthcare claims case study.
Frequently asked questions
Does the Consumer Duty apply to AI used in insurance claims?
Yes. The Duty applies to retail insurance activities in scope of ICOBS, including claims handling. Firms remain responsible for consumer outcomes when they use AI-supported tools, including products, services, communications and support delivered through AI.
Can AI decline a claim on its own under the Consumer Duty?
A solely automated decline that produces a legal or similarly significant effect engages UK GDPR Article 22, which restricts that processing unless an exception and safeguards apply, including a route to human intervention. Separately, the Duty expects good outcomes, fair support and meaningful human oversight of higher-impact decisions. Many firms keep a person on declines for both reasons.
What evidence does the FCA expect for AI in claims?
Boards need outcomes monitoring, not process completion alone: defined good and poor outcomes, metrics that show what customers actually experience, analysis for vulnerable groups, and a causal chain from poor outcome to investigation to remediation. For AI specifically, the FCA also points to testing across consumer characteristics, bias and accuracy controls, explainability, and live monitoring.
How should AI treat customers in vulnerable circumstances in claims?
FG21/1 expects firms to understand vulnerability needs, equip staff to respond, design flexible support, and monitor whether vulnerable customers get outcomes as good as others. AI journeys that exclude vulnerable customers by default, or make it hard to reach a person, can create foreseeable harm under the Duty.
Is Consumer Duty the same as the EU AI Act for UK claims teams?
No. The Duty already applies to UK retail insurance claims. The EU AI Act is a separate regime that reaches some UK insurers only where EU law applies. For how the Act maps to UK insurers, see our EU AI Act guide; for how claims processing, handling and adjudication differ, see our AI claims processing explainer.
References
The sources below were opened and checked for this review on 28 October 2026. Bracketed numbers in the text point here.
- Financial Conduct Authority, PS22/9: A new Consumer Duty (27 July 2022). fca.org.uk
- Financial Conduct Authority, FG22/5: Final non-Handbook Guidance for firms on the Consumer Duty (PDF). fca.org.uk
- Financial Conduct Authority, FG21/1: Guidance for firms on the fair treatment of vulnerable customers. fca.org.uk
- Financial Conduct Authority, "AI in financial services: good practice and areas for improvement" (published 19 October 2026). fca.org.uk
- Financial Conduct Authority, Artificial Intelligence (AI) update – further to the Government's response to the AI White Paper (22 April 2024; PDF also available). fca.org.uk
- Financial Conduct Authority, Insurance multi-firm review of outcomes monitoring under the Consumer Duty (26 June 2024). fca.org.uk
- Financial Conduct Authority Handbook, ICOBS 8 Claims handling. handbook.fca.org.uk
- Information Commissioner's Office, Automated decision-making and profiling (UK GDPR individual rights guidance). ico.org.uk
- Information Commissioner's Office, "What is the impact of Article 22 of the UK GDPR on fairness?" (AI and data protection guidance). ico.org.uk
- UK GDPR Article 22 (retained EU law as published on legislation.gov.uk). legislation.gov.uk
Last reviewed: 28 October 2026.
Not legal advice. This article is a plain-English summary for general information, written by a technology founder rather than a lawyer. How the Consumer Duty, ICOBS, UK GDPR and related guidance apply to a specific insurer depends on its facts. Take advice from a qualified lawyer before making compliance decisions.