)
At Money20/20 Europe this June, our CTO Carlos Candeias joined leaders from across banking, payments, fintech and technology for the Banking Innovation Roundtable. The session set out to explore a question every financial institution is now grappling with: how do you move from AI experimentation to enterprise-scale adoption that creates real value?
The conversation surfaced a clear consensus. AI experimentation is everywhere, but translating successful pilots into measurable business outcomes remains hard. The barrier is rarely the technology. It is data quality, organisational mindset and the slow work of changing how people operate. As Carlos put it, a pilot proves the technology; production requires changing how people work.
For us, though, the most important thread of the discussion pointed somewhere more fundamental than adoption curves. It pointed to a shift in the very shape of financial infrastructure.
From APIs to agents
For the last decade, the question every business asked a financial provider was simple: do you have an API? Open banking, embedded payments and multi-currency accounts were all built on that expectation. APIs became mandatory infrastructure.
We believe the next question will be different: “Do you have AI connectors?”
Model Context Protocol (MCP) is the emerging open standard that lets AI assistants interact directly with software systems, rather than relying on a human to operate an interface. Crucially, it is not tied to any one provider. It is a connection standard, which means your chosen AI assistant, whichever one your business already uses, can connect to a service that supports it and act on your behalf.
That is the shift we are building toward. The aim is not for VFX to push its own agents onto you. It is for VFX to become the financial service that your agents can connect to: initiating payments, checking balances, running treasury actions directly through the AI assistant you already trust, with authentication, approval controls and a full audit trail sitting underneath. Getting those controls right is precisely why this is a direction of travel rather than a switch to be flipped.
“Today we ask vendors about APIs. In the future we’ll ask about MCP support. Organisations that expose their services to AI agents will have a significant advantage.” Carlos Candeias, CTO, VFX Financial
This is why we see MCP as the future of embedded finance. Embedded finance has always been about putting financial products where they are needed, inside the software and workflows businesses already use, rather than forcing customers into yet another app. Open banking was an early form of it. Deposit accounts and payments are part of it too. The logical endpoint is that finance becomes something agents consume directly, and the standalone finance app will matter less and less. We are working toward a VFX where our clients can offer and access financial services through the AI assistants they choose, not just through screens.
Compliance as a competitive advantage
If agentic infrastructure is the destination, the journey runs straight through compliance, and that is where the second big opportunity lies.
Fabio Mineo, CEO Europe at VFX Financial, challenged a status quo that too many in the industry still accept. “Yesterday I spoke with a customer for whom onboarding still takes eight to ten weeks on average,” he says. “Is that really how this should work in 2026? Controls are still applied manually, one by one. AI is changing how the industry operates, and compliance should be a competitive advantage, not a bottleneck.”
The shift Fabio describes is from compliance as a one-time gate to compliance as a continuous, real-time process. With richer metadata connectivity and AI-driven monitoring, institutions can underwrite and monitor risk far more dynamically. Instead of excluding entire customer verticals because they are hard to assess, providers can apply granular, ongoing risk controls to serve them safely. That is not just better risk management. It is access to markets that some of our clients have previously written off.
Global from day one, fragmented underneath
There is a related friction that almost every modern business feels. Platforms are global from day one, operating across multiple jurisdictions and currencies almost as soon as they launch. Yet treasury remains stubbornly fragmented, stitched together across banks, regions and manual processes.
An embedded treasury layer, delivered through API and increasingly agent-ready connectors, can take that operational friction off the customer’s plate. Multi-currency, multi-jurisdiction treasury workflows that once demanded constant manual intervention become something that runs quietly in the background. For a scaling business, that is the difference between treasury being a tax on growth and treasury being an enabler of it.
What we took away
The roundtable reinforced something we already believed: this is not really an AI transformation. It is an operating model transformation. The technology is advancing quickly and will keep doing so. The real differentiator over the next five years will be how fast organisations can change the way work gets done while maintaining trust, governance and accountability.
Increasingly, that means designing not just for human users, but for AI agents acting on their behalf. Embedded finance, real-time compliance and connected treasury are three expressions of the same underlying shift, and they are the directions we are building toward.
If you are thinking about how agentic infrastructure, faster onboarding or embedded treasury could change the way you serve your own customers or finance operations, get in touch, we'd be keen to hear from you and compare notes.