Sage Intacct’s May 2026 R2 release includes a structurally significant addition: an AI Gateway built on the Model Context Protocol (MCP), enabling third-party AI assistants — Anthropic’s Claude, Microsoft Copilot, Cursor — to connect directly to ERP financial data via a read-only, OAuth 2.0-secured REST API (EisnerAmper, 2 June 2026). The feature is available in the United States, United Kingdom, Canada, Australia, and South Africa (Compare the Cloud, May 2026).
Context
The Model Context Protocol is an open standard designed to normalise how AI tools connect to business systems. Sage describes it as a USB port for AI: a single, structured entry point into financial data that requires no custom integration work on the client side (EisnerAmper).
In practice, the AI Gateway provides read access to general ledger, accounts payable, accounts receivable, treasury, purchasing, and order management modules. Existing Intacct permissions — roles and approval workflows — apply in full, meaning AI assistants can only surface data the authenticated user is already entitled to see.
Running alongside the Gateway, Sage is deploying a Finance Intelligence Agent (Sage Copilot) capable of answering natural language questions against live Intacct data. This module is currently in an Early Adopter programme, with broader availability planned across the rest of 2026 (EisnerAmper).
The announcement builds on earlier momentum: in April 2026, Sage and PwC had already introduced an agentic AI delivery model for Intacct, targeting faster implementation timelines and greater explainability in AI-driven results.
Impact for businesses
For IT leaders: extension without re-platforming. The core appeal of the MCP approach is that it requires no changes to the ERP itself, and no proprietary connectors. A CIO can plug any MCP-compatible AI assistant into financial data for conversational reporting, ad-hoc analysis, or decision support — without opening a new integration programme. This is AI-as-a-layer logic, not ERP replacement.
For CFOs: querying data without the BI intermediary. The Finance Intelligence Agent targets a specific use case: letting finance teams ask natural language questions about accounting, cash flows, or outstanding payables — with the system providing reasoning alongside each answer. Jon Fasoli, SVP of Sage Intacct, frames the strategic direction clearly: high-performing finance teams are no longer just closing and reporting — they are managing cash and anticipating outcomes (Compare the Cloud).
For governance: a framework to establish now. Read-only access and role-based permission enforcement are sound technical guardrails. But organisations activating the AI Gateway will need to define what users are permitted to query via LLMs against financial data, which outputs can feed into formal decision processes, and how requests are logged. The EU AI Act reinforces this requirement for high-risk systems: even where Sage operates the gateway infrastructure, usage responsibility remains with the customer.
What to watch
The Finance Intelligence Agent is still in Early Adopter status. The next critical milestone will be general availability and the first real-world reliability reports on natural language queries applied to live accounting data. Equally important: whether other mid-market ERP vendors adopt MCP as a shared interoperability standard, or each maintains a proprietary format — an outcome that will determine the portability of AI investments on the customer side.
For broader context, see our comparison of agentic AI in ERP: SAP Joule, Sage Copilot, Odoo, and Microsoft and our analysis of EU AI Act compliance for ERP programmes.
