Supply'd AI agents Bring Expense Agent and MCP to Food ERP

Supply'd, the ERP platform built for food manufacturers, wholesalers and retailers, has expanded its artificial intelligence suite with version 5.5 of Supply'd Intelligence, a release that puts Supply'd AI agents directly inside the operational workflows food businesses already use. According to the company announcement, reported by GlobeNewswire on October 28, 2026, the update introduces Ask Supply'd, a private-beta Expense Agent and read-only Model Context Protocol connectivity for compatible AI assistants.

The release matters because food businesses run on a constant stream of small operational problems. Supplier prices change, perishable inventory moves fast, purchasing and production requirements shift, and documents pile up in high volumes. Supply'd AI agents are designed to work inside that reality rather than sitting alongside it as a separate chatbot, answering questions from authorised business data and, in the case of the Expense Agent, investigating problems and recommending fixes.

Ask Supply'd answers questions from inside the ERP

Ask Supply'd, part of Supply'd Intelligence and launching in beta, lets food manufacturers, wholesalers and retailers ask questions inside their ERP and receive answers based on their own business data and Supply'd help documentation. According to the announcement, the assistant respects existing user permissions and is read-only in its initial release, so it cannot alter operational records.

That restraint is a deliberate design choice, and it reflects a wider pattern in enterprise agent deployment. The first version of an agent that can see real business data is often read-only on purpose: it proves it can find the right answer and respect permissions before it is trusted to change anything. Supply'd AI agents follow that path, with Ask Supply'd answering first and the more action-oriented Expense Agent operating behind an approval step by default.

The approach also fits how food teams actually work. A question about a supplier, a price, a delivery or a production run is rarely a general knowledge question. It is a question about this business, this week, with these numbers. Supply'd AI agents answering from inside the ERP, using authorised data, avoid the copy-and-paste loop of exporting reports into a separate tool and hoping the model has the right context.

Expense Agent investigates invoices and recommends fixes

Expense Agent, launching as a private beta with selected customers, goes further by reviewing supplier invoices, investigating discrepancies and recommending corrective actions. According to the announcement, depending on the issue, it can update invoice lines, correct tax details, schedule price changes or request credits. Actions require approval by default, while businesses can selectively permit automatic approval as they gain confidence in the system.

Invoice reconciliation is a telling first job for Supply'd AI agents. A mismatch between what was ordered, what arrived and what was billed is small on its own, but across hundreds of suppliers and thousands of lines it becomes a major drain on time. Founder Jason Stockton described that burden in the announcement, noting that individually these issues seem small, but together they consume an enormous amount of time. An agent that can investigate the discrepancy, rather than merely flag it, changes the job from chasing paperwork to approving a proposed resolution.

The approval default matters just as much as the capability. Food businesses cannot afford an agent that quietly rewrites invoices or tax details without oversight. By requiring approval first and allowing automatic approval only selectively, Supply'd AI agents keep a human in the loop while still removing the investigative work that slows teams down. That governed path to autonomy is becoming a familiar pattern across the agent economy, as suppliers learn that trust is earned in stages.

MCP connectivity opens the ERP to outside assistants

Model Context Protocol connectivity also arrives in beta, allowing users to access authorised Supply'd ERP information through supported AI assistants. According to the announcement, MCP connections are initially read-only, with future write capabilities planned behind a control plane designed to manage permissions and authorised actions.

MCP is the open standard that lets AI assistants connect to external tools and data sources in a consistent way, and its arrival in a vertical ERP is a sign of how quickly it is spreading beyond developer tools. For Supply'd customers, Supply'd AI agents mean an outside assistant they already use could, with permission, read authorised ERP information without a custom integration for every tool. For the wider agent ecosystem, it is another example of a business system opening a governed door to agents instead of blocking them at the boundary.

That contrast is worth noting. Recent coverage of banks restricting third-party agents from online banking shows one response to agent access: close the door over fraud and liability concerns. Supply'd AI agents take the opposite approach for their own domain, opening a read-only door first and planning write access behind a permission control plane. Both responses are really about the same question, namely how much an outside agent should be allowed to see and do, and Supply'd AI agents answer it with staged permissions rather than a blanket refusal.

What version 5.5 changes for food businesses

Supply'd Intelligence builds on the platform's existing intelligent purchasing, production planning, invoice processing and delivery optimisation capabilities. Version 5.5 extends that direction while also improving expense reconciliation, adding a sell-by-unit and invoice-by-weight workflow for products such as meat, seafood and cheese, and introducing OAuth authentication for the Supply'd API, according to the announcement.

Taken together, the release describes a platform moving from software that records what happened to software that helps resolve what is happening. Ask Supply'd handles the questions, Expense Agent handles the invoice exceptions, and MCP connectivity lets other assistants join in under permission controls. None of those pieces replaces the ERP itself. Supply'd AI agents sit on top of the sales, purchasing, inventory, production, warehousing, quality, delivery and business intelligence functions Supply'd already connects in one system, according to the company.

The beta staging is also a signal to watch. Ask Supply'd and MCP connectivity will launch in beta, while Expense Agent will initially be available to selected private-beta customers, according to the announcement. How quickly those betas widen for Supply'd AI agents, and whether customers switch on automatic approval for routine invoice fixes, will say a lot about how ready food businesses are to hand routine operational work to agents. For now, Supply'd AI agents arrive with a clear pitch: AI that understands the business, can help get things done, and stays under control over what it can access and what actions it can take.

Sources: GlobeNewswire announcement from Supply'd and Supply'd. Related coverage on this site includes Fibr AI Agent Experience Optimization and Jedify Data Agents.