NatWest Group is set to trial what it calls an industry-first, fully generative audio-visual Spending Insights tool, making it the first bank to test a real-time voice-to-voice conversational AI experience for discussing personal finances, initially through Royal Bank of Scotland.
Customers will be able to talk through their finances using voice or text, with the tool responding dynamically to follow-up questions and changes in topic as the conversation moves along, turning transaction data into personalised, interactive insights meant to help spot potential savings. It was built by NatWest’s Chief AI Research Office and runs on the bank’s own proprietary Small Language Model, developed in-house rather than licensed from an outside vendor.
Dr Maja Pantic, NatWest’s Chief AI Research Officer, tied the launch to a broader philosophy about pacing AI adoption around trust rather than speed. “The most exciting AI developments are still ahead of us. Success will not be determined by who adopts AI fastest, but by who adopts it most thoughtfully. Our role is to understand emerging technologies and carry out cutting-edge research that addresses real customer needs so we can earn the trust that allows innovation to scale. We’re moving beyond simple written question and answer interactions towards more natural, audio-visual conversations that allow customers to engage with complex information in a way that feels intuitive and interactive. That’s a significant step forward in how people can interact with their bank in the future.”
Alongside this, NatWest has launched a Fraud Triage Agent inside Cora, its digital assistant. Customers worried about a suspicious transaction can now describe what happened in their own words and get routed straight to the right fraud, scam, or dispute team, instead of sorting the issue into a category themselves. Nearly 19,000 customer conversations have already been handled through it since launch, a response NatWest ties to UK Finance data showing around 80% of fraud activity now starts online or over telecoms platforms.
These launches accompany NatWest’s AI Adoption Report, drawn from research among over 2,400 NatWest customers benchmarked against 1,800 UK consumers. 57% of customers say they’re using AI noticeably more than they were six months ago. The most common uses are finding deals and products (42%), getting savings tips (38%), and checking for scams and fraud (36%). But even as AI use grows, trust still comes down to people: 81% of customers said having access to a real person remains the most important factor in trusting AI with their finances.
Solange Chamberlain, NatWest’s Retail CEO, positioned the launches as complementing human judgement rather than replacing it. “Customers are already using AI as part of their everyday lives, and that is raising expectations of what great banking should feel like. They want experiences that are simpler, more personal and more responsive, while still knowing there’s someone to turn to when it matters. Through these new tools, we’re strengthening customer relationships by combining the strength of AI with the expertise and judgement of our colleagues to help customers make sense of their money, access support more quickly and navigate important financial decisions with greater confidence.”
For everyday customers, the practical change is simple. Asking about spending or reporting a suspicious transaction should feel more like a conversation than filling out a form. What’s more telling is the tension sitting underneath it. Customers clearly want AI’s convenience, but the same survey shows they still want a person available when it counts. That’s really the balance both tools are trying to strike at once.
Herald View: The fraud tool is the more immediately useful piece, nearly 19,000 conversations already handled says more than any voice-AI demo could. But the Spending Insights trial matters longer term precisely because it runs on NatWest’s own model rather than a licensed one, a quiet signal of real investment in in-house AI capability. The number that should give banks a pause isn’t the adoption stat, it’s that 81% of customers still want a human backstop before they’ll trust AI with their money. That’s the real constraint every bank’s AI rollout will have to design around.