Finance software has long been an archipelago of disconnected islands — each tool solving its own problem, none of them speaking the same language. Spendesk wants to be the shipping lane that finally connects them.

At Money20/20 Europe in Amsterdam, Quentin Vigneau, Spendesk’s Chief Product Officer, laid out a vision that goes well beyond expense reports and invoice management. The company is placing itself at the centre of what many in the industry now call the agentic era of finance, where AI systems don’t just surface information but begin to take action on behalf of the teams they serve.

Finance software as connective tissue

Spendesk already helps businesses track expenses, manage invoices, oversee subscriptions and keep company spending within policy. Its newest direction, however, is less about replacing those workflows and more about rewiring how finance teams interact with the data those workflows produce.

Rather than logging into a reporting interface, users would put questions directly to an AI assistant — and the assistant would pull accurate, live spend data in response. The new Model Context Protocol offering announced at the event is designed to make exactly that possible, allowing finance professionals to query Spendesk data through interfaces like Claude and other large language model tools.

Vigneau frames the shift in deliberately simple terms: each piece of software a finance team uses is an island with its own logic and interface. Large language models are the shipping lanes between those islands. The goal isn’t to build yet another island but to make movement between them effortless and intelligent.

One data point underscores how fast that movement is already happening: spending on AI tools among Spendesk’s customer base has grown more than fiftyfold since 2022.

Beyond productivity: the smarter spending thesis

Most AI conversations in enterprise software start and end with efficiency — fewer manual steps, faster approvals, less time on reconciliation. Vigneau accepts those gains but is more interested in a second category he calls “spend smarter.”

The distinction matters. Working more efficiently means doing the same things with less effort. Spending smarter means using AI to surface patterns and signals that help decision-makers get more value from every pound or euro they commit. It is the difference between automating a process and augmenting a judgement.

Spendesk’s product roadmap reflects that division. Near-term MCP capabilities focus on retrieval — letting teams ask questions and get answers without leaving their AI interface. Longer-term development points toward AI agents that can support purchasing and payment processes directly, operating under controlled conditions that keep compliance intact.

LLMs are the glue that connects every tool in the finance stack and multiplies the value between them. — On agentic infrastructure for finance teams

Trust is the rate-limiting factor

Vigneau is candid about where adoption slows down. Finance teams have grown reasonably comfortable using AI for research, analysis and reporting — tasks where a wrong answer can be caught and corrected. The calculus changes the moment AI moves closer to approvals, payments and compliance-sensitive decisions. There, trust is the constraint, not capability.

His position is that solving the trust problem is fundamentally a technology and design challenge, not just a change-management one. Placing a large language model on top of existing systems without rebuilding the governance layer around it is a shortcut that finance leaders will not accept for long. What they need instead is robust auditability, configurable guardrails and clear human oversight at every point where the system could initiate an action rather than simply report one.

That framing shapes how Spendesk is staging its own rollout. Agentic capabilities arrive gradually, anchored to retrieval and read-only actions first, with write and transact capabilities following only once the trust infrastructure is in place.

Humility as product philosophy

Spendesk reached profitability while continuing to invest in AI-led finance capabilities — a combination that gives Vigneau the latitude to move deliberately rather than chase every new model release. The approach he describes is one of deliberate restraint: know when to automate, know when to surface an interface, and know when to simply give the team the right information and step back.

He uses the word “obsessed” to describe the company’s focus on its own humility as a platform — an acknowledgement that the goal is to enable finance teams to do what they want, not to make them dependent on Spendesk’s particular interpretation of what good finance looks like.

That disposition sits at the heart of the agentic bet. The companies most likely to earn finance teams’ trust as AI agents begin handling real transactions will not be the ones who moved fastest. They will be the ones who built the governance layer first and let capability follow.