Every poorly designed AI-enabled customer journey eventually meets someone it wasn’t designed for: someone in financial distress, recently bereaved, someone managing a cognitive impairment, or just someone having a bad day.
For the FCA, how that moment is handled defines whether the organisation is Consumer Duty compliant or not.
The instinct is to treat this as a technology problem: better prompts, smarter routing, a more capable chatbot. However, the evidence suggests that AI doesn’t fail vulnerable customers, poorly designed processes do.
When designing customer journeys, businesses often aim for efficiency and test against the ‘average’ customer. The problem is, according to the FCA, almost half of UK adults show at least one characteristic of vulnerability, so the businesses that aren’t testing their AI-enabled journeys against vulnerable customer scenarios – and ArvatoConnect’s research shows 73% aren’t – are neglecting a significant portion of their customer base.
Waiting for complaints about failures before addressing these customer needs make little business sense. The whole customer journey should be designed to consider every scenario where possible.
Empowered advisors are effective safeguards
Our research found that 78% of vulnerable customers actively seek some sort of human interaction when speaking to AI agents, but only 32% want a solely human-centred approach; 46% are looking for a mix of technology and human and 22% prefer technology only.
And this preference changes depending on the scenario. A customer recently told me about having to contact his bank because he was struggling to make payments. He found the situation embarrassing and preferred using an AI chatbot to access the information he needed and begin resolving the issue rather than explaining his circumstances over the phone. In other words, a human solution isn’t always the most empathetic solution.
The key point is customers want choice. The best protection against bad outcomes is an advisor with the authority to escalate to a specialist or switch channel if it better suits the needs of the customer. That authority has to be built into the operation deliberately, not left to individual initiative.
Outcome-focused quality assurance catches what SLAs miss
SLA compliance and average handling time confirm the customer was moved along – it says nothing about whether the customer received an appropriate outcome.
At ArvatoConnect, our quality assurance process is focused on outcomes rather than procedure. We make sure our agents have understood the customer’s circumstances, needs and delivered the right outcome, not whether a certain set of steps were followed.
We allow as much time as is needed to deliver the right outcome, encourage flexibility with contact channels and empower advisors to speak up if a process isn’t working for customers so it can be reviewed
Flexibility is a resourcing decision
You cannot bolt on flexibility ionto a rigid model, as it starts further upstream than an individual interaction, it’s a resourcing decision built into the operating model. At ArvatoConnect, we believe 70/30 is the optimal digital-to-human split, with human capacity concentrated where emotional intelligence and sensitive conversations are most likely to surface.
AI’s role is to create the capacity for judgement
At ArvatoConnect, AI is there to help advisors do their jobs better, accelerating the speed of finding the right information, not to replace the advisor. Used this way, AI absorbs routine volume queries, so advisors have the time and attention to apply judgement where it is needed.
AI does not change the principles of customer service 101. We get customers the outcomes they deserve, whether the process is 100% delivered by AI or a human advisor.
