Infor's New Claude Connector Can Execute Manufacturing Changes

Infor and Anthropic announced Infor Industry AI for Claude on November 12, 2026, a connector that lets manufacturing employees query operational data and execute approved transactions without leaving a Claude conversation. The initial rollout is narrow. Access is limited to select Infor Velocity Suite customers, and each customer needs its own Claude license plus an administrator to activate the connector.

The stakes are straightforward. Enterprise AI assistants have mostly been read-only: they answer questions, summarize documents, and stop there. This integration crosses that line in a controlled setting, letting Claude carry out actions inside the systems that actually run factories, warehouses, and purchasing. Infor also designated Claude as its preferred AI assistant for interacting with its products, and Claude models as the preferred models for its Industry AI solutions.

That makes the deal more than a chatbot bolted onto an ERP screen.

What can Claude actually do inside Infor?

The connector targets production planners, buyers, and customer service representatives working in either Claude or Infor. Users can review Infor data and recommendations, then authorize changes such as updating a purchase order or addressing an at-risk production order.

The examples Infor and Anthropic lay out put investigation and execution in the same exchange:

• A production planner can ask why an order is at risk, examine material shortages, limited machine capacity, or labor shortages, and then initiate corrective action.
• A buyer can see which production or customer orders a late receipt would affect, identify alternative suppliers, and place replacement purchase orders in one guided conversation.
• A customer service representative can check availability, warehouse locations, and production options before rushing an order or confirming a delivery date — while the customer is still on the line.

With approval, Claude can also reallocate stock to an urgent order, shift production lines after unplanned downtime, or ship an order. The design goal is to compress the handoffs between identifying a disruption and recording the response in Infor.

"Manufacturing depends on expertise spread across many systems, and a large part of the day goes to moving information between them," said Lauren Collett, Anthropic's head of enterprise industries GTM. The connector is built to cut that movement by bringing Infor data and actions into the same Claude exchange — without requiring customers to build and maintain a custom integration.

How does the technical bridge work?

The connector is bidirectional rather than read-only. Infor's product documentation describes an Infor Model Context Protocol server that brings operational data into Claude and writes approved actions back into Infor applications.

Model Context Protocol, or MCP, is the open standard that lets the connector work across Infor and other compatible tools. It's the same plumbing pattern other enterprise vendors are adopting to let AI assistants touch live systems rather than just read from them. Users can still switch to Infor's role-based workspaces for more detailed work, so the conversation layer sits on top of the existing applications rather than replacing them.

The technical significance is worth naming plainly. A write-back path into ERP-grade manufacturing software is the harder half of enterprise AI adoption, and it's the half most vendors have avoided because errors move physical goods, not just text.

Who controls what Claude can change?

Permissions, not goodwill, are the control layer here. Single sign-on and role-based entitlements inherited from existing Infor permissions determine what each user can see and do. Actions taken through the connector remain subject to authorization controls and are recorded in Infor. The plugin is designed to require explicit user approval before each change.

The connector also recognizes the user's identity, role, and site, so responses reflect that person's operational context. Infor configures its shared manufacturing context once for the company rather than teaching it separately to each user. Its vocabulary treats item, SKU, and product as equivalent terms, and the supplied context can include bills of materials and production orders.

Those controls create what amounts to a supervised write-back path. Claude can recommend a next step and execute it, but only after authorization from someone whose Infor entitlements permit the action. That arrangement ties Claude's ability to act to the same identity and permissions already used inside Infor — rather than treating access to the assistant as blanket permission to change any record.

Why the rollout is limited — and what comes next

Infor intends to add roles, skills, and applications over time, extending the integration beyond its initial workflows. The company hasn't published a timeline for that expansion. The phased approach fits the risk profile: writing approved changes into production and inventory systems is exactly where an unsupervised AI error would be most expensive.

Infor CEO Kevin Samuelson framed the underlying logic in a statement: "When AI begins managing your workflows, the system underneath matters more, not less." It's a pointed remark for a moment when some AI vendors pitch assistants as replacements for the software beneath them. Infor's position is the inverse — the assistant is only as trustworthy as the entitlements, audit trails, and process controls of the system it acts on.

For manufacturers, the immediate implication is a shorter path from an operational question to a recorded transaction, with approval and existing permissions still separating Claude's recommendation from a change to production, inventory, or customer orders. Whether that path shortens work or just relocates the approval clicks will depend on how the select Velocity Suite customers use it — and on whether Infor's promised expansion of roles and applications keeps the same permission model intact.