← All posts News · Nov 16, 2026

What Claude for Manufacturing means for your back office

Lennard Kooy··6 min read
What Claude for Manufacturing means for your back office

Claude for Manufacturing is a plugin Anthropic launched on 12 November 2026 with 17 skills for manufacturers, including quoting a new job and sourcing from another supplier. It covers a slice of the back office, connects to 2 ERPs and runs on Claude only. For manufacturers it is proof that this work is worth automating.

At Lleverage we think the launch means 3 things. The work is now a proven category. A back office needs one product across its processes. And the model, the ERP connection and the place the data runs should stay the manufacturer's choice. We build AI agents that run the whole back office for companies that make, move and sell physical products, and you can book a demo to see one run inside your ERP.

What did Anthropic announce?

On 12 November 2026 Anthropic released Claude for Manufacturing for Claude Team and Enterprise plans. It has 17 skills and 4 connectors: Autodesk Fusion, Infor, Microsoft 365 and Oracle NetSuite. Systems without a connector work through regular exports saved to Microsoft 365.

Anthropic names 6 example skills and calls them a starting point: quoting a new job, investigating a failed inspection, diagnosing a breakdown, sourcing from another supplier, comparing designs in Autodesk Fusion and reporting on production. It cites a 2025 Manufacturing Leadership Council survey in which 79% of manufacturers enter data manually. A person reviews and approves the work before it reaches a customer or supplier.

Infor released its own connector the same day. From Claude, a user can create a purchase order, reallocate stock to an urgent order or ship an order, and the connector is designed to ask for explicit approval before each change. It follows the user's existing Infor permissions and is available to select Infor Velocity Suite customers, who need their own Claude licence.

What do we think it means for manufacturers?

  1. The work is proven. Anthropic and Infor put quoting, sourcing and purchase orders in front of manufacturers on the same day, and both build in a person's approval. A manufacturer no longer needs convincing that an AI agent can do this work. The open questions are which one, and where it runs.
  2. A back office needs one product. A quote depends on the price agreement, the purchase order on the supplier's confirmation, the invoice on both. We think these processes belong in one product, so the customer terms, the approvals and the audit trail sit in one place.
  3. The manufacturer should keep the choices. Which model runs each step, which ERP the agent works in and where the data is processed carry a price and a risk. We think they should stay with the manufacturer.

Which part of the back office does it cover?

Of the 6 named skills, 2 are back-office work: quoting and sourcing. The other 4 are quality, maintenance, engineering and plant reporting. None of the named skills covers invoices, collections, customer service or master data. A back office runs all of those at once, on the same facts.

Does Claude for Manufacturing connect to my ERP?

The plugin has connectors for 2 ERPs, Infor and Oracle NetSuite. Changing records is described only for Infor, through Infor's own connector. Anthropic names no other ERP, so SAP, Microsoft Dynamics 365 Business Central, Exact and AFAS customers work from exports. An agent working from an export sees the ERP as it was at the last export.

Who controls the model and where the data runs?

Claude for Manufacturing runs on Claude, and Infor calls Claude its preferred AI assistant. That puts one provider's prices, outages and release schedule under the back office.

Anthropic lists EU hosting for Claude on AWS Bedrock and Google Vertex AI, with Microsoft Foundry coming soon. It does not say what applies to the Team and Enterprise plans that Claude for Manufacturing installs on, so ask before you sign. Orders, customer prices and supplier terms are the sensitive part of a manufacturer's business.

Who decides what the agent does alone?

Anthropic and Infor both put a person's approval in the flow. In Rootstock's 2026 survey of manufacturers, 73% prefer human-directed AI or AI that works on its own within set limits and escalates, and only 13% want fully autonomous AI. Rootstock sells manufacturing ERP, so read it as vendor research.

Approval before every change keeps a person in control. We think routine steps with clear rules should run on their own once the numbers earn it, with a person keeping the exceptions.

Why run this on Lleverage?

Lleverage is one product for the whole back office, and each gap above has a concrete answer.

  • It covers quotes and orders, purchasing and supplier confirmations, customer questions, invoice matching and collections, forecasting and planning and master data, with a full audit trail.
  • It runs inside the ERP you already have. The agent fills the same fields a person fills, through the system's own interfaces, so the ERP checks the work the same way. It connects to SAP, Business Central, Dynamics 365, Exact, AFAS, Infor, NetSuite and others. In August 2026, between 45 and 71 customer purchase orders a week were written into Microtechniek's on-premise Ridder iQ without typing.
  • You choose the model. Pick it for the agent and for each workflow step, from providers including Anthropic, OpenAI and Google, or connect your organisation's own provider account.
  • You choose the region. Lleverage is EU hosted by default and GDPR compliant, holds SOC 2 Type II and ISO 27001, and keeps zero retention.
  • You set how much runs alone. The agent scores how sure it is of each piece of work. Below the line a person decides, above it the work goes through on its own, and only the customer moves the line.
  • The first process typically runs supervised within a few weeks, built by our forward deployed engineers with your team.

Claude for Manufacturing and Lleverage can sit side by side: Claude for engineering and quality questions, Lleverage for the back office.

What should a manufacturer do now?

  1. Count the orders, confirmations and invoices that still start with someone typing.
  2. List the processes you want covered and ask whether you are buying one tool per process or one product.
  3. Ask each vendor which of your systems it changes, not only which it reads.
  4. Ask which models it runs on and whether you can change them.
  5. Ask where the data is processed and what is retained.

Start with one process and let the agent earn more. Our guides on AI order processing for manufacturers and back-office automation for manufacturers show where to begin, and a demo shows how it runs on your ERP.

Frequently asked questions

What is Claude for Manufacturing?

A plugin Anthropic launched on 12 November 2026 for Claude Team and Enterprise plans. It has 17 skills, among them quoting a new job, investigating a failed inspection and sourcing from another supplier, and 4 connectors: Autodesk Fusion, Infor, Microsoft 365 and Oracle NetSuite. A person approves the work before it reaches a customer or supplier.

Does Claude for Manufacturing connect to my ERP?

It has connectors for 2 ERPs, Infor and Oracle NetSuite. Infor's own connector can change records, designed to ask for approval before each change, and is available to select Infor Velocity Suite customers. Other systems, including SAP, Business Central, Exact and AFAS, are not named. They can be used through regular exports saved to Microsoft 365.

Which AI is best for manufacturing?

It depends on the step. Reading a messy PDF, checking a pricing rule and drafting a customer reply differ in the accuracy, speed and cost they need. We think a manufacturer should choose the model per step and change it later, instead of tying the whole back office to one provider.

Can AI handle a manufacturer's back office, not just the shop floor?

Yes, for repetitive work with clear rules: reading orders in any format, drafting quotes, handling purchase orders and supplier confirmations, matching invoices and answering customer questions with order context. A person decides the exceptions. Most AI agents for manufacturing, such as maintenance and scheduling agents, work on the shop floor, which is a separate problem.

Can a manufacturer keep its data in the EU with Claude?

Anthropic lists EU hosting for Claude on AWS Bedrock and Google Vertex AI, and lists Microsoft Foundry as coming soon. It does not say what applies to the Team and Enterprise plans that the manufacturing plugin installs on. Ask Anthropic or your reseller where inference runs, where data is stored and what is retained.

Why choose Lleverage over Claude for Manufacturing?

Claude for Manufacturing is an assistant a person starts, on one model, with 2 ERP connectors. Lleverage covers orders, purchasing, invoices, customer questions and master data as one product, inside the ERP you run, on the model you choose, hosted in the EU. The two can sit side by side, Claude for engineering questions and Lleverage for the back office.

Give your back office an AI workforce.