By Joseph Jack, Head of Compliance, ArvatoConnect
Artificial intelligence is transforming financial services at pace. But while much of the conversation focuses on technology, models and automation, the real question organisations should be asking is much simpler: are we delivering better outcomes for customers?
At ArvatoConnect, we believe responsible AI isn’t about finding more places to deploy technology. It’s about understanding where customers struggle, designing services around their needs, and using AI to remove friction rather than create it.
For financial services organisations, this matters more than ever. Whether someone is buying a home, managing debt, recovering from a difficult life event or facing financial uncertainty, these are some of the most important moments in a person’s life. For vulnerable customers, the stakes are even higher. The way organisations design customer journeys – with or without AI – can have a lasting impact on trust, confidence and financial wellbeing.
The gap between AI ambition and customer reality
Our latest research, Care by Design, surveyed 1,000 senior decision-makers across UK financial services organisations alongside 1,000 financially vulnerable customers.
The findings reveal a clear contradiction.
On one hand, organisations are embracing AI with confidence. Almost nine in ten (88%) have increased their use of AI in customer-facing operations over the past year, and the same proportion believe AI has the potential to improve outcomes for vulnerable customers. Businesses recognise the opportunity to deliver faster support, personalise customer journeys, simplify communications and help advisers work more effectively.
Yet confidence falls sharply when organisations assess their own readiness.
Only 23% believe their AI strategy is not increasing the risk of digital exclusion or biased outcomes for vulnerable customers.
At the same time, nearly three-quarters (74%) of financially vulnerable customers admit they have felt like giving up while trying to get help from their bank, insurer or financial provider.
That disconnect should concern every organisation.
AI has enormous potential to improve customer experience, but if organisations prioritise technology over customer need, they risk scaling poor experiences just as effectively as good ones.
Consumer Duty starts with customer focused design
Many organisations are understandably focused on where AI can improve efficiency. The bigger challenge is ensuring it improves customer outcomes.
Our research suggests many businesses are still missing this opportunity.
Only 31% sandbox test AI systems for biased or unethical outcomes before deployment. Just 27% test against real vulnerable customer scenarios, while only 26% carry out formal impact assessments focused specifically on vulnerable customers.
The result is that AI is often introduced before organisations fully understand how it will behave in real customer environments.
The answer isn’t necessarily more regulation.
Instead, organisations should return to the fundamentals of good customer service.
That starts with understanding customers and defining what good outcomes actually look like. Once those outcomes are clear, services can be designed around them. Before deployment, solutions should be tested against realistic scenarios, including complex and sensitive customer journeys where mistakes carry greater consequences.
Most importantly, the work doesn’t stop once technology goes live.
Customer feedback, outcome monitoring and continuous improvement should become part of the operating model. Good service design is never static.
This approach closely reflects the principles of Consumer Duty. The objective isn’t simply to demonstrate AI adoption. It’s to demonstrate that customers consistently receive fair, effective and accessible support.
Technology should support judgement, not replace it
Responsible AI isn’t about removing people from customer service. It’s about helping them make better decisions.
If we take those design principles we’ve just talked about – understanding customers, designing for outcomes, and continuously testing and improving – the next question is: what does that actually look like in practice?
When we started looking at where AI could genuinely improve outcomes, we didn’t start with the technology. We started with the customer journey.
We looked for journeys where the opportunity to improve both customer and business outcomes was greatest – particularly where interactions were complex, customers could be vulnerable, and the cost of getting it wrong was high.
One area stood out very clearly: complaints handling.
Across the financial services industry, around 3.6 million complaints are raised every year, with over £500 million paid in redress. But what we saw wasn’t simply an operational challenge driven by volume. It was a customer outcome challenge.
Customers weren’t always receiving prompt responses. Outcomes weren’t always consistent because cases required complex judgement. And in some situations, customers didn’t feel their individual circumstances had been fully understood.
So we asked a simple question: How can we deliver fair, consistent, explainable and faster outcomes, while keeping people at the centre of the decision?
That led us to work with XpertRule to develop our Augment Complaints Engine.
The solution combines Symbolic AI with Generative AI to support advisers with structured, evidence-based guidance, improving speed, consistency and compliance while keeping humans fully in control of final decisions.
Unlike traditional black-box AI models, our deterministic, “glass-box” approach makes decision pathways transparent and explainable. The information is controlled, the reasoning can be traced, and organisations remain accountable for every outcome – something that is becoming increasingly important in regulated industries.
Most importantly, technology supports the adviser rather than replacing them.
AI provides structure, consistency and access to information at speed, while the adviser brings empathy, context and judgement. It’s that combination that delivers better customer outcomes.
And that reflects one of the clearest findings from our research. Customers want choice. They’re happy to use AI for simple, transactional requests, but when situations become sensitive, emotional or complex – particularly for vulnerable customers – they still want a human in the loop.
Ultimately, responsible AI isn’t about replacing people. It’s about combining technology with human expertise to deliver better outcomes for every customer.
Building trust into AI
One of the clearest findings from our research is that customers still want choice.
For routine, transactional interactions, many are happy to use self-service tools or digital assistants.
But when situations become complex, emotional or involve vulnerability, people still want human support.
That’s where trust is built.
AI should enhance customer journeys, not define them. It should reduce effort, improve consistency and free advisers to focus on the interactions where empathy and judgement matter most.
Customer service 101
Responsible AI isn’t really about artificial intelligence.
It’s about customer service.
The organisations that will succeed won’t necessarily be those deploying the most advanced technology. They’ll be the ones that understand their customers best, design journeys around real outcomes, continuously test and improve those journeys, and embed governance throughout the process.
AI absolutely has the potential to transform financial services. It can make services faster, simpler and more accessible.
But if organisations lead with technology rather than customer need, they risk creating services that are operationally efficient while failing the very people they’re intended to support.
Customer service 101 still applies. Put people first. Technology should follow.
