Consumer Duty Unreasonable Barriers: Insurance Phone Guide

A claimant rings to report a burst pipe and is asked for the same policy details three times before anyone takes the claim. A policyholder whose father has died calls to close his car insurance and is told the attorney holding his affairs must write in. A customer who wants to complain is routed through a menu that offers renewals, new quotes and payments, and nothing that says complaint. None of these calls breaks a script. Each one sits inside the consumer support outcome of the Consumer Duty, and each one now has to survive a question an insurer should be able to answer: was that step necessary, or was it a barrier?

This guide is written for the Head of Customer Operations at a UK personal lines insurer who owns the telephone estate, whether it runs on human agents, an outsourced partner or voice AI. It covers the consumer support rules in PRIN 2A.6 of the FCA Handbook, which have applied since 31 July 2023, and how a phone journey can be designed to remove unreasonable barriers while keeping the friction the rules actually require. Its scope is deliberate. Auto-renewal cancellation parity under ICOBS 6A.6 is covered in depth by our voice AI guide for insurance broker renewals, collections conduct by the debt recovery Consumer Duty guide, detecting vulnerability by the vulnerable customer detection guide, and claims data capture by the AI voice claims intake guide. This post holds the altitude above them: the whole support journey, the seven moments the rule names, and the line between a barrier and appropriate friction.

This guide is shipped by the team behind Dilr Voice, a multi-agent voice AI platform that chains specialised agents into a single phone call. Or see DATS, the five-stage AI consulting system that maps where AI belongs in a regulated operation.

What does the Consumer Duty mean by an unreasonable barrier on a phone journey?

Under the FCA Consumer Duty, an unreasonable barrier is a step that pushes a retail customer into unreasonable extra effort to get something done. PRIN 2A.6.3G names steps that are unreasonably onerous or time consuming, complex to carry out, or difficult to understand, plus requests for unnecessary information or evidence. On an insurer's phone line, that covers menus, queues, repeated security questions and demands for paperwork the firm does not need.

The rule itself sits in PRIN 2A.6.2R, and it is a rule, not guidance. A firm must design and deliver support so that it meets the needs of retail customers, including those with characteristics of vulnerability, lets them use the product as reasonably anticipated, includes appropriate friction to mitigate harm, and ensures they do not face unreasonable barriers, including unreasonable additional costs, during the lifecycle of a product. The guidance in PRIN 2A.6.3G then says those additional costs include unreasonable exit fees or other charges, delays, distress or inconvenience. Distress and inconvenience matter for a phone journey, because a long and confusing call costs the caller something even when no fee is charged.

Two features of the application provision, PRIN 2A.6.1R, decide who owns the problem. The section applies regardless of the channel used, so a phone line, a webchat and a voice agent are all in scope on the same terms. And it applies where the firm outsources its interactions with retail customers to a third party, in whole or in part. The duty binds the insurer. It does not move to the outsourced contact centre or to the vendor whose software answers the call, which is why the design decisions in this guide belong to the insurer's own AI operating model rather than to a supplier's default settings.

Which seven moments does PRIN 2A.6.2R(4) name, and how do they map to insurer calls?

PRIN 2A.6.2R(4) lists seven moments where a retail customer must not face unreasonable barriers: making general enquiries or requests, amending or switching the product, transferring to a new provider, accessing a benefit the product is meant to provide, submitting a claim, making a complaint, and cancelling or terminating the relationship. For a UK insurer, almost every one of those seven arrives by telephone at least some of the time.

Mapping them to call types is the first useful exercise, because it shows which calls a contact centre has historically treated as cost and which it has treated as revenue. The table below is our own mapping of the rule's seven moments to typical personal lines calls. It is a design aid, not FCA text.

The value of the table is that it forces the same question across all seven rows. A contact centre that has redesigned its claims line can still leave the complaint journey untouched, and the rule treats both the same. Our guide to where AI pays in UK insurance covers the FCA's work on claims handling and board reporting; consumer support is the thread that runs through both.

What does the FCA guidance say about call waiting times?

The FCA guidance in PRIN 2A.6.4G says a firm would be unlikely to meet its consumer support obligations if its support causes prospective customers to be prioritised over existing customers, or causes unreasonable delays when existing customers try to engage. It names call waiting times directly, which makes queue design a direct Consumer Duty question for an insurer's telephone operation.

The exact words are worth reading slowly, because they are guidance with a conditional lead-in, not a flat ban. PRIN 2A.6.4G(2) describes support that causes "unreasonable delays when retail customers attempt to engage with the firm, including disproportionately longer call waiting times to cancel or make changes to an existing product than to purchase a new product", as quoted from the FCA Handbook, PRIN 2A.6. The test is disproportion, not equality. A cancellation queue a few seconds longer than the sales queue is not the target; a cancellation line that routinely runs far longer than the sales line, because sales gets the staff, is.

The same guidance goes further than queues. PRIN 2A.6.4G(3) names unreasonable delays to payments due after they have been agreed, to the firm requesting necessary information or evidence, and to the firm processing information or evidence it has received. For an insurer, that third limb reaches past the phone line into claims operations: a voice agent that takes a claim in four minutes and then waits a week to ask for the photographs it needed has not removed the barrier, it has moved it. Our guide to voice AI callback and virtual queue design covers the mechanics of queue management; the Consumer Duty question is which journeys get priority within it.

The FCA has also tested this in practice. Its review of the consumer support outcome, published 7 March 2025 and last updated 10 July 2026, reported instances of longer wait times for existing customers looking to terminate or downgrade their product, and said some firms need to consider how they make it at least as easy to switch out of a product, leave or make a change as it is to buy in the first place.

How is appropriate friction different from sludge?

Appropriate friction is a deliberate step that protects a customer from foreseeable harm and gives them a real chance to understand their options. Sludge is a step that exists to stop them acting in their own interests. The FCA Consumer Duty requires the first in PRIN 2A.6.2R(3) and treats the second as an unreasonable barrier, so an insurer has to be able to say, for every step on a phone journey, which one it is.

PRIN 2A.6.3G(2), which is guidance, addresses the overlap: where a firm has included appropriate friction to comply with PRIN 2A.6.2R(3), that friction would not amount to an unreasonable barrier. The FCA's consumer support review is candid that the boundary is a judgement for the firm, describing sludge as practices "that restrict customers from acting in their interests" and asking firms to weigh whether good outcomes are likely and foreseeable harm can be mitigated. That puts the burden of reasoning on the insurer. A step that cannot be justified by a named harm is hard to defend as friction.

The review gives an example, from a firm whose sector it does not name, that shows the judgement working in both directions. One firm introduced digital chat journeys for customers looking to cancel or downgrade a policy, and in doing so carried out risk assessments that considered situations where a customer is at risk of under-insurance and a telephone conversation may be appropriate. The FCA called this a positive example of a firm considering the effect of its support channels on customers' ability to exit products. A phone journey can apply the same logic, and more of our insurance and regulated-sector writing works through similar judgements.

A useful working test is to write down, for each step, the harm it prevents. Confirming that a caller who wants to cancel buildings cover understands that a mortgage lender may require it is friction tied to a harm. A second confirmation call, a mandatory retention script before the instruction can be accepted, or a request for a reason when no reason is needed, usually has no harm attached. The FCA review also observed that some firms appear to prioritise making sales or retaining customers over good customer outcomes. The review did not address retention scripts specifically, but a retention step with no stated harm is hard to square with that observation.

How should a voice agent remove barriers without removing protection?

A voice agent removes barriers when it is designed around the customer's objective rather than the firm's preferred outcome. For an insurer, that means naming every one of the seven moments as a route the caller can reach, asking only for the information the step genuinely needs, giving cancellation and complaint calls the same priority as sales, and handing over to a person whenever friction is required.

The same diagnostic logic underpins our AI placement diagnostic, which ranks call types by where an agent removes the most barrier for the least risk before any build starts.

Six design decisions carry most of the weight, and each maps back to a provision.

1. Name every moment at the first prompt. A caller who says "I want to complain" or "my mother has died" should be understood in their own words, not forced through a menu that lists only commercial options. This answers PRIN 2A.6.3G(1)(a)(iii) on steps that are difficult to understand.
2. Ask for what the step needs, once. Verification should be proportionate to the risk of the change, and data already captured should travel with the call. On Dilr Voice, each specialised agent passes context to the next on handover, so the next agent does not start from nothing.
3. Route by need, not by value. Cancellation, complaint and bereavement calls should not sit behind sales in the queue. This is the PRIN 2A.6.4G(1) and (2) test.
4. Make friction a handover, not a wall. Where a conversation protects the customer, such as possible under-insurance, the agent should explain why and offer a warm transfer to a person with full context, which our warm transfer guide covers in detail.
5. Accept the instruction. Once the customer has the information friction is meant to give them, the instruction is recorded and confirmed. In our view, a loop that keeps re-offering a save, with no harm attached, fails that test.
6. Record the journey. Every call should leave an audit trail showing what was asked, what was offered and how long each step took, so the insurer can evidence the design rather than assert it.

None of these decisions is specific to AI. A human contact centre can fail every one of them, and the FCA's review findings were about firms of all kinds. What changes with a voice agent is that the design is explicit: the routes, the questions and the handover rules are written down before the first call, which makes them testable in a way a training manual rarely is. Our enterprise guide to AI voice agents sets out how those agent types and handovers fit together.

The diagram below sets out the order in which we would take an insurer's phone estate through the exercise. It encodes this guide's own method, not an FCA process.

What do authorised representatives and other firms need from an insurer's phone line?

PRIN 2A.6.5R requires an insurer to give a person authorised by a retail customer or by law, such as someone holding a power of attorney, the same level of support it would give the customer. PRIN 2A.6.6R separately requires the firm to deal with reasonable requests from another firm effectively and in good time. Both are rules, and both are easy to break on a phone line built only for policyholders.

The representative rule is where many voice journeys fall short, because identity and verification flows are written for one person. A caller acting under a power of attorney, or a family member dealing with an estate, often cannot answer the security questions the policyholder would answer. A journey that simply fails verification and ends the call gives that person a lower level of support than the customer, which the rule does not allow. A better design recognises the representative early, routes them to a path that checks their authority rather than the customer's memory, and treats them as a customer for every step after that. If the verification flow is the part an insurer is least sure of, it is a good place to speak to the operators before a build starts. There is one limit in the text: PRIN 2A.6.5R(2) says the requirement does not apply where the person assisting is also a firm.

That is where PRIN 2A.6.6R takes over. Requests from another firm, such as a new provider asking for policy history or a broker acting on the customer's behalf, must be dealt with effectively and in good time so the other firm can support the customer. On the phone, that argues for a recognised route for firm-to-firm requests and a service standard for answering them, rather than leaving them in the general queue.

Representatives and bereaved callers are often dealing with difficult circumstances, and the FCA review gives a reference point for how firms said they support customers with characteristics of vulnerability. Its quantitative survey of retail financial services firms, run in May 2024, received 356 responses from 407 firms surveyed across banking, insurance, payments, consumer finance and investments.

The figures are shares of firms in a cross-sector survey, not a measure of insurers alone, and they describe what firms said they had rather than whether it worked. They do not measure representative or bereavement routes at all. They are still a useful prompt for an insurer: if 48% of firms in the survey reported dedicated telephone support for customers with characteristics of vulnerability, it is worth checking whether the representative and bereavement routes on its own phone estate were designed on purpose or inherited.

How should an insurer evidence that its phone journeys carry no unreasonable barriers?

An insurer evidences the consumer support outcome by monitoring what happens to customers on each journey, not just how fast calls are answered. The FCA's 2025 review found some firms relied on transactional metrics, such as contact rates and wait times, or waited for complaints to surface problems. For a phone estate, the stronger evidence compares journeys with each other and follows the outcome past the end of the call.

Three comparisons give a Head of Customer Operations most of what a board report needs. The first is the gap between journeys: time to reach a person, time to complete the instruction and number of steps, measured separately for buying, changing, claiming, complaining and cancelling. Disproportion is a comparison, so it can only be evidenced by one. The second is repeat contact within a short window on the same issue, which shows whether a call actually resolved anything. The third is the outcome of the handover: when a voice agent transfers a cancellation or complaint, whether the instruction was completed.

The FCA's review also describes some firms using proactive tools, such as speech analytics in call centres, to assess the quality of support, and suggests, in its section on smaller firms, that monitoring could focus on higher-risk journeys such as bereavement, financial difficulty or cancellations, while noting that this does not remove the need for monitoring to identify poor outcomes across the whole customer base. That is a sensible order for a first quality programme, not a limit on it, and our voice AI agent quality scoring guide covers how to build the scorecard.

Outsourcing changes the evidence question rather than removing it. The review found mixed progress in firms implementing effective management information oversight where the customer support function is outsourced to a third party. Because PRIN 2A.6.1R applies to outsourced interactions, an insurer that runs its overflow through a partner, or its first line through a voice agent, still needs journey-level data back from that operation. A contract that delivers only service level figures does not supply it, and building that reporting into the weekly operating rhythm is the kind of work our AI execution office runs.

The customer side of the evidence is not reassuring. The FCA's Financial Lives 2024 survey, as cited in the same review, found that in one in five recent contacts or attempted contacts with a financial services provider, finding the right contact information was very or fairly difficult, or not possible. In 13% of contacts where the customer got through, what the provider said was very or fairly difficult to understand. Those figures cover financial services generally, but they point at two barriers an insurer's contact design influences: getting to the right place, and understanding what happens there. Our voice AI governance guide for financial services sets out how this evidence fits the wider supervisory picture.

What is the best voice AI approach for Consumer Duty phone journeys in 2026?

The best voice AI approach for Consumer Duty phone journeys in 2026 is the one that lets an insurer write down and test its routes, questions and handover rules, give exit journeys equal standing with sales, and produce journey-level evidence for the board. Dilr Voice supports parts of that pattern, but an existing contact-centre platform is sometimes the better place to fix the problem.

The criteria matter more than the brand. An insurer should ask whether the system recognises all seven moments in the caller's own words, whether verification can branch for representatives, whether a handover carries full context to a person, whether every call leaves an audit trail, and whether the insurer, not the vendor, controls queue priority. Dilr Voice offers inbound front desk handling with business-hours configuration and after-hours routing, warm transfer to a human with full context, opt-out recognition, and full audit trails on every call, and its product page sets out the agent types. The best AI voice agent 2026 buyer guide compares the wider field.

There is an honest concession. If the barrier is that cancellation calls sit behind sales in the routing plan, the fastest fix is usually a change to that routing in whatever contact-centre platform the insurer already runs, such as Genesys or Five9, not a new voice agent. A voice agent, whether from PolyAI, Dilr Voice or another vendor, is worth evaluating where the barrier is in the conversation itself: menus that do not name the request, repeated questions, and handovers that lose context. An insurer that cannot yet say which kind of barrier it has should map the journeys first, which is the work our DATS methodology starts with.

Frequently asked questions

Does the Consumer Duty ban retention offers on insurance cancellation calls?

The FCA Consumer Duty does not name retention offers in PRIN 2A.6, so an insurer is not banned from mentioning an alternative on a cancellation call. The test is whether the step is necessary. Under PRIN 2A.6.3G, a step that makes cancelling unreasonably onerous or time consuming, or that asks for unnecessary information, is an unreasonable barrier, and a retention script that must be completed before the instruction is accepted should be tested against that standard.

The FCA's review also noted that some firms appear to prioritise retaining customers over good outcomes, which an insurer should keep in mind when designing any save step. An insurer should be able to show where a single, clearly optional offer ends and a barrier begins.

Does the Consumer Duty still apply when an insurer outsources its phone line?

Yes. The FCA Consumer Duty support rules in PRIN 2A.6 apply to an insurer's interactions with retail customers including where the insurer outsources those interactions to a third party, in whole or in part. The duty stays with the insurer whether calls are answered by its own staff, an outsourced contact centre or a voice agent, so the insurer needs journey-level data back from every party that handles its customers.

Where a voice agent runs the first line, part of the evidence should come from the agent's own records. Dilr Voice keeps a full audit trail on every call, which the insurer can draw on alongside its own journey data.

Is a longer wait to cancel than to buy automatically a breach?

Not automatically. The FCA guidance in PRIN 2A.6.4G says a firm would be unlikely to meet its consumer support obligations if its support causes disproportionately longer call waiting times to cancel or change an existing product than to buy a new one. The word that matters is disproportionately, so an insurer should measure the gap between the two journeys, explain any difference, and act when the gap grows without a reason tied to customer harm.

For auto-renewing general insurance, the separate ICOBS 6A.6 rules add specific requirements on cancellation methods, which our broker renewals guide works through in detail.

Want to see this in production? Try Dilr Voice live, book an AI placement diagnostic, compare options in the best AI voice agent 2026 guide, or read about our approach to placing AI inside regulated operations.

Written by the Dilr.ai engineering team, practitioners who ship enterprise AI in production. Follow us on LinkedIn for shipping notes, or subscribe via the RSS feed.