Skip to content

How UK Credit Unions Can Use Conversational AI Without Losing the Personal Touch

DigitalWell
DigitalWell
How UK Credit Unions Can Use Conversational AI Without Losing the Personal Touch
9:32

UK credit unions served 2.29 million members in 2025, according to the Bank of England's annual statistics, with total assets of £4.99 billion. Many of them run with small teams, and many members value the fact that someone answers the phone and knows their name. That's the tension in any conversation about AI: the efficiency is real, and so is the risk of losing what makes a credit union different.

 

UK credit unions can use conversational AI without losing the personal touch by automating the routine, factual requests members want answered fast, keeping people on anything involving hardship, judgement or vulnerability, and making it easy to reach a person at any point, in line with the FCA's Consumer Duty.

TL;DR

  • Members value speed on simple things and a person on difficult ones. Conversational AI should split the work along that line.
  • Balance queries, payment status, card blocking and opening hours suit automation. Arrears, hardship, bereavement and complaints stay with people.
  • The FCA's Consumer Duty and its guidance on vulnerable customers mean AI must support good outcomes and never block access to help.
  • Credit unions are dual-regulated by the PRA and FCA, so outsourcing, operational resilience and data protection all shape the choice of provider.

What do credit union members actually value?

Members value two things that pull in different directions: quick answers to simple questions, and a real person when something is difficult. A member checking whether their loan payment has gone through at 9pm doesn't want to wait until the morning. A member who has just lost their job and can't make that payment wants to talk to someone who understands.

 

Conversational AI can serve the first need very well. It can undermine the second if it's put in the way. The design question is how to give members the speed without taking away the person.

Which member requests should AI handle, and which should stay with people?

Draw the line by the kind of request. The table sets out a starting split.

 

Request

Suitable for AI?

Why

When to hand to a person

Balance and recent transactions

Yes

Factual, high volume, needs live data

If the member queries a transaction they don't recognise

Loan payment status

Yes

Factual, often out of hours

If the payment has failed or the member mentions difficulty paying

Lost or stolen card

Yes, to block quickly

Time-sensitive, clear process

If fraud is suspected on the account

Opening hours and branch details

Yes

Simple, repetitive

Rarely

Loan application status

Yes

Factual, reduces chasing calls

If the member wants to discuss the decision

Arrears or payment difficulty

No

Needs judgement, forbearance options, empathy

Immediately

Bereavement

No

Sensitive, requires care and discretion

Immediately

Complaints

No, beyond logging

Regulated process, needs a person

Immediately

Signs of vulnerability

No

Consumer Duty and FCA guidance on fair treatment

Immediately

 

The right-hand column matters as much as the left. An AI that recognises a trigger, such as a missed payment or distress in the member's words, and passes the member to a person quickly is protecting the personal touch rather than replacing it.

How does Consumer Duty shape the design?

The FCA's Consumer Duty, in force since 31 July 2023, requires firms to deliver good outcomes for retail customers. One of its four outcomes is consumer support: customers should be able to get help easily, without unreasonable barriers.

 

For conversational AI, that has practical consequences. Members must be able to reach a person without going round in circles. The AI must not make it harder to complain, cancel or ask for help. And the credit union should monitor whether AI-handled members get outcomes as good as those who speak to staff.

 

The FCA's guidance on the fair treatment of vulnerable customers adds to this. Firms should be able to recognise signs of vulnerability and respond appropriately. For AI, that means clear triggers for handing over, and staff who receive the handover with enough context to help.

What other rules apply to UK credit unions using AI?

Credit unions in Great Britain and Northern Ireland are authorised by the PRA and regulated by both the PRA and the FCA. That brings three further considerations when choosing an AI provider.

 

Outsourcing and operational resilience. An AI service answering member calls is an outsourced function. Treat it like any other: ask for the provider's continuity arrangements, incident process and exit terms, and check the PRA and FCA requirements that apply to your credit union's outsourcing.

 

Data protection. Call recordings, transcripts and chat logs are personal data under UK GDPR, regulated by the Information Commissioner's Office. Know where they're stored, for how long, and on what lawful basis.

 

Transparency. Tell members when they're speaking to an automated assistant. Beyond good practice, it supports the Consumer Duty's expectation that communications are clear. Credit unions with members in the EU should also check the EU AI Act's disclosure duty.

How should the handoff to a person work?

The handoff is where the personal touch is won or lost. Four things make it work.

 

  1. Offer a person at any point. A simple phrase, such as "speak to someone", should always work.
  2. Pass the context. The staff member should see who the member is, whether they've been verified, and what they've already said.
  3. Route to the right person. Arrears go to the team that handles arrears; a lost card with suspected fraud goes to someone who can act on it.
  4. Keep the member informed. If there's a wait, say how long, and offer a callback instead of silence.

 

Ross Murray, CEO and Co-Founder of DigitalWell, describes the experience DigitalWell brings to regulated deployments:

 

"We're able to talk to years of complex deployments within highly regulated industries." Ross Murray, CEO and Co-Founder, DigitalWell

What does conversational AI look like on the phone?

For many members, the phone is still the main way to reach their credit union. Voice AI at the network layer can answer those calls without the credit union replacing its phone system.

 

DigitalWell's AI Voice Network runs AI Voice Agents inside its voice network. They verify members, answer balance and loan status queries, block lost cards and give branch hours, then hand over with a summary when a person is needed. Every call is transcribed, summarised and logged, which also supports the monitoring Consumer Duty expects. DigitalWell provides Ireland, UK and international numbering, so members keep ringing the number they know.

 

DigitalWell works with credit unions in Ireland, where its Credit Unions page sets out the calls AI handles and the supplier documentation it provides. The same principles apply to UK credit unions, with the PRA and FCA in place of the Central Bank of Ireland.

How should a UK credit union start?

Start small, measure member outcomes, and involve the people who answer the phone today.

 

Pick one or two routine requests, such as balance queries and card blocking. Agree the handover triggers with the staff who deal with vulnerable members. Tell members what's changing and how to reach a person. Then compare outcomes for AI-handled and staff-handled calls over the first months, and report them to the board as part of Consumer Duty monitoring.

 

Staff can become strong supporters once routine calls move off their desks, because they get more time for the conversations members remember.

Frequently Asked Questions

Does Consumer Duty allow credit unions to use AI with members?

Yes. Consumer Duty doesn't prohibit AI. It requires good outcomes and easy access to support, so AI must be designed to help members and never block them from reaching a person.

Should AI ever handle arrears or payment difficulty calls?

No, beyond recognising the situation and passing the member to a trained person quickly. These calls need judgement, forbearance options and empathy.

How do we spot vulnerable members if AI answers the call?

Set clear handover triggers, such as mentions of hardship, illness or bereavement, and missed payments. Review a sample of AI-handled calls regularly to check the triggers work.

Do we need to replace our phone system?

No. Network-layer voice AI works with the existing phone system and numbers.

Is conversational AI realistic for a small credit union?

Yes, if it's limited to a few routine requests at first. Small teams can gain the most, because the same people cover the counter, the phone and the back office.

Key Takeaways

  • Split the work by request type: AI for fast factual answers, people for hardship, judgement and vulnerability.
  • Consumer Duty's consumer support outcome means members must reach a person easily at any point.
  • Handover triggers, full context and the right team are what keep the personal touch intact.
  • PRA and FCA outsourcing expectations, UK GDPR and clear AI disclosure shape the choice of provider.
  • Start with one or two routine requests, and compare outcomes for AI-handled and staff-handled members.

 

Exploring AI for your members' calls? Talk to DigitalWell about starting with one routine request and a handover design your staff agree with.

Share this post