Claude for Manufacturing Targets Plant Paperwork, Not the Machines
Anthropic is aiming Claude at the paperwork around plant operations, not the machines. It cites a 2025 survey in which 79% of manufacturers said they still enter data by hand, and its new Claude for Manufacturing plugin for Team and Enterprise plans brings 17 skills and four connectors covering sales, quality, operations, supply chain, design and setup. Some skills handle a single task; others chain several steps until a person must review.
The examples are the unglamorous work that eats a plant’s office hours. Quoting a new job means reading the customer request, listing gaps, comparing costs on similar jobs and drafting the quote. Investigating a failed inspection means documenting it, checking past reports, ranking likely causes and drafting the corrective action report. Other skills diagnose breakdowns from sensor exports, repair notes, photos and manuals, prepare supplier requests for quotes and compare bids, generate and score design variations in Autodesk Fusion, and compile production reports showing output, downtime and its cost.
Connectors reach Autodesk Fusion, Infor, Microsoft 365 and Oracle NetSuite; systems without connectors can be used through exports saved to Microsoft 365. Infor is releasing its own plugin so Claude can also make changes there — releasing a purchase order to another supplier, or rescheduling production when a steel shipment slips — within the buyer’s Infor permissions. Every skill ends with human review, Claude links figures back to source documents, and it does not operate equipment.
The design lines up with where Deloitte says industrial work is heading. A Deloitte Insights article argues classroom training is failing factory-floor workers and points to AI knowledge agents delivering knowledge at the point of work, including micro-lessons built within 48 hours of near-miss incidents. The common thread is putting the answer where the work happens.
For mid-market manufacturers, the adoption question is narrower than “should we use AI.” Pick the quote, the corrective action report or the downtime report that costs the most hours today, run it through the skill with review in place, and measure whether turnaround and review time fall.