AI Is Moving Into the Systems That Run Factories: Why the Anthropic–Infor Partnership Matters

FutureIsNow Editorial
9 Min Read
Two industrial robotic arms on an automated assembly line in a factory.

AI agents are moving beyond answering questions to executing approved business workflows. Manufacturing could become one of the first major tests of whether enterprises are ready to trust AI with operational decisions.

For years, artificial intelligence in manufacturing has largely been associated with predictive maintenance, quality inspection, robotics and production analytics. These applications help manufacturers identify inefficiencies, anticipate equipment failures and improve decision-making. But a more consequential shift is beginning to emerge: AI agents in manufacturing are moving closer to the enterprise systems that actually run factories and industrial businesses.

On November 12, 2026, Anthropic introduced Claude for Manufacturing, a specialized offering designed to help manufacturing teams work across existing software systems. The announcement coincided with a partnership with Infor, enabling users to initiate approved actions within enterprise applications through Claude.

The significance is not simply that employees can ask AI questions about factory operations. It is that AI is moving closer to the systems where procurement, inventory, production planning, supplier management and customer orders are coordinated.

From answering questions to taking action

Anthropic says Claude for Manufacturing includes 17 manufacturing-focused skills and four connectors covering Autodesk Fusion, Infor, Microsoft 365 and Oracle NetSuite. The intended workflows include preparing quotations, investigating quality issues, reviewing production data, comparing designs and working with supplier information.

Infor’s integration takes the proposition a step further. Its Industry AI for Claude plugin is designed to let users initiate approved actions in Infor applications from a conversational interface. Examples described by the companies include creating purchase orders, reallocating inventory, adjusting production plans after downtime and coordinating order fulfilment.

That distinction matters. A system that summarizes a production report is an assistant. A system that can prepare and initiate a change to an enterprise record is moving into the territory of an operational agent. Infor says the integration works with existing role-based permissions, authorization controls and audit records. Availability is initially limited to select Infor Velocity Suite customers, with administrator activation and a separate Claude license required.

Why the enterprise software layer matters

Manufacturing businesses do not operate through isolated applications. Their operations depend on interconnected systems that coordinate materials, suppliers, employees, machines, production schedules and customer commitments. Enterprise Resource Planning (ERP) software sits at the center of much of this activity.

A delay in receiving a critical component, for example, can affect several production orders, delivery commitments and customer relationships. Understanding the consequences requires information from multiple functions, not a single spreadsheet or dashboard.

This is where industry-specific context becomes important. A general-purpose model may understand what a purchase order or inventory shortage means. But assessing the operational consequences requires access to relevant business records, manufacturing logic, organizational permissions and process rules.

If AI agents become a primary interface through which employees interact with enterprise applications, the value of the underlying software may increase rather than disappear. The ERP system still holds the operational data, business rules and permissions. AI changes how people access that information and act on it.

The shift also connects to a broader change in the AI infrastructure stack: as models become more capable, the differentiator increasingly lies in the systems, data and networks that make them useful in real business environments.

From copilots to operational agents

A copilot typically assists a person with a task: summarizing information, drafting documents, answering questions or generating recommendations. An agent is designed to work through a sequence of steps towards a defined outcome. It can retrieve information, assess options, coordinate actions and interact with connected systems within the permissions it has been granted.

Consider a factory facing an unexpected shortage of a critical component. Employees may need to check supplier records, examine inventory across locations, identify affected production orders, assess alternatives and coordinate revised delivery schedules. An agent connected to the relevant enterprise systems could bring those steps together, identify the consequences and prepare possible responses for approval.

The potential benefit is not simply faster information retrieval. It is reduced friction between identifying a problem and responding to it. But this should not be confused with autonomous factory management. Anthropic says Claude for Manufacturing does not operate physical equipment; the announced capabilities focus on information, analysis and workflows that remain subject to human review.

Trust could become the biggest barrier to adoption

Allowing AI to summarize a production report is relatively low risk. Allowing it to initiate a purchase order, change inventory records or affect a production schedule introduces a different level of responsibility.

An incorrect recommendation can be reviewed and discarded. An incorrectly executed action could affect suppliers, customers, production schedules and financial commitments. This makes governance central to enterprise agent adoption.

Infor says its integration uses single sign-on, role-based permissions and existing authorization controls, with actions recorded in its systems. These safeguards matter because an agent should not automatically receive broader access than the employee using it.

Yet permissions alone will not resolve every challenge. Enterprises will need to evaluate recommendation reliability, data quality, exception handling, reversibility and accountability. They will also need to establish whether productivity gains justify the cost of integration, implementation, licensing and ongoing oversight.

The announcements establish new capabilities, not independently verified productivity gains or large-scale manufacturing return on investment. Those outcomes still need to be demonstrated.

What this means for India’s manufacturing and technology ecosystem

India has an opportunity to participate in this transition. As manufacturers modernize their operations, demand could grow for professionals who understand both industrial processes and enterprise AI implementation. This creates opportunities across industrial software, systems integration, manufacturing engineering, data architecture and AI-enabled operations.

India’s Global Capability Centres (GCCs) could also take on responsibilities involving enterprise AI development, workflow integration, industrial analytics and technology operations. The evolution of these centres is already moving beyond cost-efficient delivery towards deeper global technology mandates, as explored in our analysis of India’s next-generation GCC model.

For Indian manufacturers, the opportunity is more immediate but also more demanding. Effective deployment will depend on the quality of existing digital infrastructure, integration of operational data and employee readiness. Companies with fragmented data, outdated software and poorly documented processes may find that deploying agents is considerably harder than installing a plugin.

The competitive advantage may therefore belong to organizations that strengthen their digital foundations before attempting to automate complex decisions.

The next enterprise AI advantage

The Anthropic–Infor partnership is an early example of a broader direction in enterprise technology: AI models are becoming more deeply integrated with the systems businesses depend on. The next phase of competition may not be determined exclusively by which company builds the most capable model. It may also depend on which platforms can access relevant enterprise context, work across existing applications and execute tasks within clearly defined governance frameworks.

For manufacturing leaders, the questions are becoming practical. Which workflows can safely be delegated to AI? Which decisions must remain with employees? How will performance be measured? And what safeguards are needed before agents interact with systems responsible for production and customer commitments?

The factory of the future may not be defined only by the number of robots on its production floor. It may also be defined by how intelligently its software systems coordinate decisions, people and resources.

The shift has begun. The real test is whether enterprises can trust AI not just to understand their operations, but to act within them.

Image credit: Simon Kadula / Unsplash.

Share This Article