Banks want what AI promises: faster resolution, lower cost and better service. What they cannot afford is to trade away customer trust or operational resilience to get it.
Cameron Thomson, group vice president for the EMEA theatre at enterprise communications company Avaya, watched regulated organisations wrestle with exactly that tension at the company’s Journeys UK event in London. His answer starts with reframing the question itself.
“Financial institutions should start by treating AI as an extension of their existing trust and resilience requirements, not as an exception to them,” he says. “The question is not simply, ‘Where can we automate’ It is, ‘Where can AI act safely and responsibly, where should a human remain involved, and how do we preserve control throughout the process?”
He offers a banking scenario demonstrated at the London event. AI identifies a suspicious transaction, blocks the card and begins resolving the issue digitally. But when the customer requests a replacement card at a different address while travelling, the risk profile changes and the interaction moves to a human specialist, who receives the full context rather than forcing the customer to start again.
“That is the model I expect financial institutions to pursue: automation where confidence is high, human judgment where the stakes rise, and resilient communications that underpin both,” says Thomson. “Trust comes from knowing AI can act quickly and within boundaries without being allowed to act indiscriminately.”
Why regulated is different
The case for that caution rests on consequences. “In highly regulated sectors, the consequences of getting AI wrong are simply more significant,” he says. “A poor retail recommendation may be annoying. An incorrect decision involving a financial transaction, customer identity or fraud alert can create severe regulatory, financial and reputational consequences very quickly.”
That changes the architecture as much as the use case, he argues. Financial institutions need to know where their data resides, who and what can access it, how decisions are governed and whether critical services remain available if another part of the technology environment fails. It is why he sees continued demand for hybrid and on-premises deployment alongside public cloud, particularly in Europe, where sovereignty is a major consideration. “For regulated organisations, flexibility is not resistance to innovation. It is often what makes responsible innovation possible.”
Context before autonomy
Thomson believes open standards will do much of the connective work, singling out the Model Context Protocol (MCP), which gives AI a standardised, governed way into systems such as CRM platforms, operational applications, knowledge sources and APIs. “A language model on its own can generate an answer, but it does not inherently know what is happening in a customer’s account right now, or have permission to take action,” he says. For a bank, that could mean securely connecting an AI agent to fraud information, customer records and approved workflows rather than building a different custom integration for every AI model or application. He adds that an AI-agnostic orchestration layer matters for the same reason: “enterprises are not forced to make a permanent bet on a single model provider as the AI
market continues to evolve.”
The payoff he describes is real-time intelligence: bringing customer, operational and risk information together while the interaction is still happening. A card suddenly used abroad can trigger the appropriate response immediately, and if the customer then asks for something that changes the risk profile, the bank can route the conversation to a specialist with the fraud alert, authentication status and prior conversation already in view. “The objective was not simply to detect fraud faster, but to move from detection to resolution without fragmenting the experience,” he says of the journey demonstrated at the event. It gives banks a chance, he adds, to make service and risk decisions from the same current picture rather than from separate silos.
Modernisation, but selectively
His strongest takeaway from Journeys UK was about the pace of change that regulated organisations will actually accept. “They are not rejecting modernisation or innovation. They are rejecting the idea that modernisation has to mean replacing everything at once, and at scale,” he says. Customers want access to AI, better orchestration and richer data, but they also have systems and integrations deeply embedded in their operations for good reason. In financial services especially, continuity, sovereignty, security and regulatory requirements cannot become secondary considerations simply because a newer technology is available.
Gradual modernisation resonated instead: existing workflows analysed with AI, translated into understandable business logic and then moved into a modern environment, improving processes along the way rather than blindly recreating them. “I think the next phase of financial-services modernisation will be much more selective: preserve what is mission-critical, open it up where appropriate, and introduce new capabilities where they deliver measurable value.”
Over the next few years, Thomson expects resilience and openness to become increasingly interconnected. “As financial institutions introduce more AI agents, applications and data sources, the underlying communications environment actually becomes more important, not less. More intelligence creates more dependencies, and those dependencies have to remain available when customers or employees need them,” he says. Nor does he expect institutions to build that future around closed technology stacks. “AI is developing too quickly. An open architecture allows an organisation to change models, introduce specialist providers and connect new systems without redesigning the entire customer journey each time.”
“Over the next few years, the strongest institutions will combine those two ideas, an open layer for rapid innovation and a resilient core they know they can depend on.”
The post How Banks can Adopt AI Without Losing Customer Trust appeared first on The Fintech Times.
