Last week, I had the pleasure of co-hosting an executive roundtable for TP alongside Intelligent Sourcing at The Rubens at the Palace in London. We brought together customer experience, risk, and compliance leaders from across banking, telco, utilities, and logistics to tackle one of the most pressing questions facing our industry: as service journeys become digital-first, how do we protect vulnerable customers without replacing empathy and accountability with pure automation?
Titled ‘The Empathy Imperative: Protecting vulnerable customers in an AI-driven world’, the evening opened with a keynote from Abhijnan Dasgupta , Practice Director Everest Group | R&D and Innovation Leaders .
Everyone involved in designing customer experiences recognises the challenge. As more customer service processes are automated, how can organisations ensure they spot when a customer needs extra support?
Vulnerability is far more common, and far less obvious, than many executives assume. Research shared by Everest showed that 49% of UK adults exhibit at least one characteristic of vulnerability, with 37% experiencing multiple overlapping drivers. Crucially, as we discussed around the table, many customers never explicitly declare that they are struggling. In many cases, they may not even realise extra help is available.
This is where AI can step in as a powerful early-warning radar. AI is exceptionally good at identifying subtle behavioural patterns, such as a customer repeatedly failing to complete a digital process, struggling with a specific step, increasing contact frequency, or suddenly altering their usual communication style.
A frontline agent handling high call volumes might miss these subtle signals, whereas AI can surface them instantly. However, as a room, we agreed on a crucial boundary: AI should never autonomously label a customer as “vulnerable,” alter terms, or deny service based on automated inferences. Instead, AI’s role is to flag potential distress, offer tailored support, and seamlessly bring a human into the interaction when judgment is required.
One of the points from our discussion that really resonated with me was that switching to “human support” does not automatically mean forcing a customer into a phone call.
Take bereavement as an example. Having to repeatedly explain a painful life event to multiple agents can be deeply distressing. For many people in that situation, a clear, step-by-step digital process offers privacy and control. Sometimes a customer needs help, but the last thing they want is a phone conversation. The real challenge is not “Humans vs. AI.” Instead, it is giving vulnerable customers genuine choice and maintaining context across channels.
During the evening, my TP colleagues and I shared how we are using AI behind the scenes to support the wellbeing of frontline agents. Sticking with the bereavement example, if our workforce orchestration system detects an agent has handled several emotionally heavy calls in a row, AI can automatically prompt them to take a rest break or temporarily reallocate them to lower-stress webchat channels without waiting for a manager to intervene. AI is not delivering or replacing human empathy; it is protecting our people so they have the emotional bandwidth to provide genuine care when it matters most.
Another key takeaway was that vulnerability is fluid. Illness, bereavement, or sudden financial pressure can happen overnight. If you rely on a static “vulnerability flag” raised years ago, your data is already outdated. Detection must be dynamic and continuous across the live journey.
The overarching message from Everest and our roundtable discussion came down to risk and harm potential. Low-risk, reversible tasks are ideal for self-service automation. But when a customer faces confusion, bereavement, or complex consent, human judgment must take over.
AI should not be used to replace customer service professionals. It should act as a radar system that spots distress before human agents can, empowering frontline teams to deliver better outcomes. For any decision that truly impacts a customer’s life, the path to human judgment must always remain open.
