Pleo has launched a suite of AI agents designed to automate routine spend management tasks for finance teams.
The announcement positions Pleo's AI capabilities as a response to what the company describes as a structural tension in finance operations: teams facing growing administrative load at the same time as pressure to operate with greater speed. According to research cited by Pleo, 49% of UK companies report that current processes slow them down, and 78% of finance professionals believe the future of the function depends on the effective use of AI - yet 68% indicate that AI skills and confidence remain underdeveloped.
From workflows to agents
According to the official press release, Pleo is introducing five specialised agents, each scoped to a distinct area of financial management. The Policy Agent, already live, enforces company spend rules in real time, flags exceptions, and routes approvals. The AP Agent handles invoice processing end-to-end, from ingestion through email to payment tracking. The Treasury Agent monitors cash flow and tracks spend against budget, surfacing overspend risk before it materialises. In addition, the Accounting Agent manages transaction coding, account reconciliation, and book closing, escalating only items requiring human review. A fifth capability, Pleo MCP, is described as the next feature to launch; it is designed to capture card transactions, match receipts, generate memos, apply accounting codes, and submit expenses automatically.
Beta testing is scheduled to begin in July 2026, with a broader rollout to follow.
Infrastructure and data access
Alongside the agents themselves, Pleo is making its MCP (Model Context Protocol) server available as an infrastructure layer. This allows finance teams to connect third-party AI tools (including ChatGPT Codex, Gemini, and Copilot) directly to Pleo's data environment. Through these integrations, users can query spend data, approve expenses, and run reports without switching between platforms.
Pleo states that its dataset, drawn from 40.000 business customers across Europe, underpins the intelligence built into the agents, providing visibility into spending patterns and risk signals that the platform has accumulated over a decade of operation.
A stated design principle for the agentic experience is that the AI is built for functional reliability rather than conversational affect, a deliberate departure from AI interfaces that, according to Pleo, replicate human emotional responses in ways that can obscure underlying issues.
On data governance, Pleo notes that customer data remains encrypted and under customer control throughout, with the MCP server architecture intended to address concerns around data use that the company says are common among finance teams evaluating AI tools.
