Vulnerability is not a niche concern in UK customer service. The Financial Conduct Authority estimates that 47% of the UK adult population has one or more characteristics that would classify them as vulnerable, whether through poor health, bereavement, financial difficulty or limited digital confidence. Technology has real potential to help this group live more independently, but it is not yet trusted to replace a person when things get complicated.
ArvatoConnect’s report, Technology and Vulnerable Customers: How to Improve Accessibility and Make Services More Inclusive, surveyed 250 UK consumers who self-identify as vulnerable and found that 74% have experienced a situation where a human would have handled things better than an AI or technology tool, and 78% still want some level of human interaction when seeking support.
There is no single agreed percentage, but the FCA’s benchmark that 47% of UK adults have a characteristic of vulnerability is the figure most contact centres use as a working estimate, and ArvatoConnect’s own survey of self-identified vulnerable consumers found that around a third have three or more overlapping vulnerabilities at once.
The FCA defines a vulnerable consumer as someone who, due to their personal circumstances, is especially susceptible to detriment, driven by four factors: health, resilience, life events and capability. ArvatoConnect surveyed 250 people who self-identify as vulnerable, evenly spread from 18 to 65-plus and evenly split by gender, to understand what that looks like day to day.
ArvatoConnect’s survey found that mental health conditions were the most common vulnerability cited (46%), followed by chronic or long-term health conditions (34%) and physical disabilities (28%). Visual or hearing impairment, age-related vulnerabilities, neurodiverse or neurological disabilities, financial circumstances and social or cultural vulnerabilities each affected 15 to 18% of respondents.
Around a third of participants reported three or more vulnerabilities at once, and the data shows real overlap: almost half of those with age-related vulnerabilities also have a physical disability or visual or hearing impairment, and mental health issues correlate strongly with chronic illness, physical disability and financial difficulty. ArvatoConnect treats multiple, overlapping vulnerabilities as the norm, not the exception, and checks for compounding needs rather than acting on the first vulnerability a customer discloses.
ArvatoConnect has built inclusive, human-backed contact services for two named UK public sector programmes: the Department for Energy Security and Net Zero’s (DESNZ) fuel-poverty support campaign, and National Records of Scotland’s first online census, both designed around digitally excluded and vulnerable citizens from the outset.
For DESNZ, ArvatoConnect advised on inclusive service design for a campaign supporting fuel-poor households, addressing limited digital literacy and reliance on carers or guardians to complete applications, with the explicit aim of eliminating barriers to accessibility so no one was excluded from seeking support.
For National Records of Scotland’s first online census, ArvatoConnect set up a contact centre to help the public complete it. Agents were trained to support people with diverse vulnerabilities, assisted by tech-based tools and specialist language services, walking respondents through the census questions and entering answers directly for anyone who was struggling.
Use AI and automation for simple, repetitive tasks and keep a fast, visible route to a human for anything complex, personal or urgent, because 78% of vulnerable consumers ArvatoConnect surveyed want at least some level of human interaction, and 74% have experienced a situation where a human handled things better than a tech tool.
ArvatoConnect’s survey found a clear, consistent split in how vulnerable consumers want to be supported: 46% want a combination of human interaction and technology, 32% prefer human interaction alone, and only 22% prefer technology alone. Respondents draw the line by task, not by preference for one channel over another: tech is welcomed for quick, routine tasks like paying a bill, while complaints, healthcare queries and anything involving money or a relapse in someone’s condition consistently pull people back toward a person.
ArvatoConnect’s survey found that trust in AI drops further for money and health specifically: respondents were largely unwilling to hand banking or medical queries to an algorithm, citing concerns about safety and a wish to be dealt with by someone who will not judge them. Simple, low-value transactions were the exception.
Test before launch and keep testing after launch, because although 76% of the vulnerable consumers ArvatoConnect surveyed had not personally felt biased or discriminated against by an algorithm, the risk of inadvertent bias, racial, gender or socioeconomic, remains real for a population already facing disproportionate disadvantage.
Only 24% of ArvatoConnect’s survey group said they had felt bias or discrimination from technology algorithms or AI. That is a reassuring number on its face, but customers are often unaware of hidden bias in the tools they use. Discrimination can show up in different forms, racial, gender or socioeconomic, and vulnerable populations already face significant disparities elsewhere, which raises the stakes of getting this wrong.
ArvatoConnect believes that, applied responsibly, AI can also help firms mitigate discrimination and reduce inequality rather than add to it. The distinction is deployment discipline. Three commitments follow from the data:
Good design combines a genuine choice of channels with AI that adapts to the individual rather than the average customer, following six principles from ArvatoConnect’s research: keep every channel truly accessible, protect postal and phone services, humanise the digital experience, keep processes current, treat AI as a supporting shoulder rather than a replacement, and properly train and support frontline agents.
ArvatoConnect’s recommendations, drawn directly from the survey, set out six principles for good design:
ArvatoConnect’s own recommendation is to treat AI as a trusted digital shoulder for both customers and agents, not a replacement for either.
Build a guaranteed, fast handover into every automated journey rather than a self-serve escape hatch, because 78% of the vulnerable consumers ArvatoConnect surveyed want some level of human interaction, and many describe being trapped in a bot loop as one of the most alienating parts of automated support.
ArvatoConnect’s survey respondents were candid about where automation fails them: chatbots that do not recognise a question however it is phrased, systems that send people in loops without resolving anything, and the frustration of eventually having to phone in anyway after wasting time on a tech-based option. ArvatoConnect builds the fix structurally: dedicated phone lines, specialist support teams and priority customer schemes that guarantee a route to a person, rather than a chatbot that occasionally offers to transfer.
ArvatoConnect is a UK customer experience and business optimisation partner that works with organisations looking to reshape how they serve customers, including some of the world’s biggest brands and UK public sector bodies. Its 2026 report, Technology and Vulnerable Customers: How to Improve Accessibility and Make Services More Inclusive, surveyed 250 UK consumers who self-identify as vulnerable about their experiences of human and digital support.
The findings sit behind three commitments ArvatoConnect builds into every client engagement: introducing technology, from intelligent automation to AI, that empowers agents to deliver personalised support rather than replacing them; being people-centric, investing in the skills, wellbeing and engagement of its own people so they can do the same for customers; and delivering resilient contact centre solutions built to withstand a wide range of threats, from cyber attacks to challenging economic conditions. for regulated environments, where decisions must be explainable and a human stays accountable for the outcome.
