Manufacturing has connected its systems but not its execution

Manufacturers have spent years connecting ERP, PLM, MES, CRM and other enterprise platforms, yet some of the most important commercial processes still depend on people moving between those systems and reconstructing what is happening. A quotation may require engineering validation and material availability to line up with production capacity before anyone can make a credible delivery promise. Once that order is accepted, the same commitment must survive changes across the business without becoming detached from the operational reality behind it.

Manufacturing Today has explored that problem in a new joint white paper with Infosys Topaz AI Next, The orchestration gap: Building the agentic manufacturing enterprise without replacing the core. The paper examines why technical connectivity has not necessarily created connected execution, and how Agentic AI could provide an orchestration layer across existing systems without forcing manufacturers into another wholesale technology replacement. The underlying argument is that the next stage of automation is less about transferring data between applications and more about coordinating decisions across them.

Connectivity is not coordination

After years of investment, manufacturers now operate sophisticated technology estates spanning engineering, production, customer management and finance, but those systems were generally designed to perform specific functions rather than coordinate an end-to-end business process. The difficulty appears when responsibility moves between functions and information must be interpreted rather than simply transferred. At those boundaries, connected applications can still leave employees responsible for understanding what has changed and deciding what needs to happen next.

Across quote-to-cash, a single customer request can move through several systems and functions before anyone can make a reliable commitment. Information may arrive through email, EDI or a customer portal, then pass through engineering and commercial checks before planning and procurement can confirm whether the request is deliverable. Each function may complete its own part successfully, yet the wider process still depends on employees spotting when an assumption has changed and working out what that means for the decisions already made elsewhere.

The white paper describes the resulting cost as a fragmentation tax. Employees spend time reconciling information and resolving mismatches between systems, while decisions made during quotation can become disconnected from the conditions that will determine whether the promise can eventually be kept. That cost appears in slower response, rework and customer commitments that become harder to fulfil as circumstances change.

Traditional integration has solved an important part of the problem by allowing data to move between applications and established workflows to automate predictable transactions. Its limitations become more obvious when a revised engineering specification affects a delivery promise or when a supplier issue forces several functions to reconsider decisions that have already been made. At that point, technical connectivity may still leave the wider process dependent on human coordination.

Agentic AI moves automation between systems

Agentic AI introduces a different model because software can begin to interpret information and coordinate activity rather than simply execute a fixed sequence. One agent might extract information from an RFQ while another validates product references, with the orchestration layer determining what should happen next and whether the case needs human intervention. The important change is that automation can begin to operate across the gaps between systems rather than stopping when work leaves a predefined path.

Manufacturers do not need an AI equivalent of every employee involved in quote-to-cash for that model to create value. The more practical opportunity is to deploy specialist capabilities around the points where people currently bridge disconnected systems, then coordinate those capabilities around an individual quotation or order. That approach allows automation to respond to changing circumstances without pretending that every decision can be reduced to a deterministic workflow.

The paper also avoids the assumption that progress depends on replacing ERP, PLM or MES. Infosys Topaz AI Next positions orchestration as an intelligent layer across the existing estate, allowing core applications to remain the systems of record while agents work across information and workflows distributed between them. For manufacturers that have accumulated multiple ERP systems or regional processes through years of growth and acquisition, that distinction could be significant because the route to better execution does not have to begin with harmonizing the entire technology stack.

Evidence from one aerospace and defense supply chain deployment provides a useful indication of what that model can deliver. Infosys Topaz AI Next reported a three percent increase in quote-to-order conversion, around a 30 percent reduction in quote-intake handling effort and approximately 60 percent fewer order-entry defects. In-stock quotations were returned in less than 12 hours, while automated workflows achieved full traceability.

Those improvements are more interesting when considered across the process rather than as isolated efficiency gains. A faster quotation creates limited value if configuration errors generate rework later, just as efficient order entry cannot compensate for unreliable fulfillment. The greater opportunity lies in keeping commercial commitments aligned with what engineering and operations can deliver as conditions change.

Automation still needs boundaries

Extending coordination across the process does not mean removing people from every decision. The white paper argues that Agentic AI may be most useful between deterministic automation and fully manual processing, particularly where exceptions still consume large amounts of employee time. Agents can assemble context and resolve part of a problem before passing the decision to a person when the consequences justify human judgment.

That makes governance part of the operating model rather than something added after deployment. Manufacturers need to decide which actions an agent can take automatically and where approval is still required, while retaining clear visibility into how decisions were reached. As workflows become more interconnected, that visibility becomes increasingly important because an error in one part of the process can influence decisions further downstream.

The more credible route to adoption is therefore unlikely to begin with agents everywhere. Manufacturers can start with tightly bounded problems such as RFQ intake or order changes, establish what good performance looks like and then extend orchestration into the surrounding process once those initial deployments are proven. Existing automation can continue to handle predictable work, while agents are reserved for the points where interpretation or coordination adds value.

Reliable product and master data remain essential wherever automated decisions depend on them, but that does not mean manufacturers have to complete an enterprise-wide data transformation before they can begin. A more practical approach is to improve and map the information required for clearly defined use cases, then extend that discipline as orchestration moves into adjacent parts of the process. This allows progress to begin around real operational needs while the wider data estate continues to mature.

The longer-term objective is not a manufacturing enterprise in which every decision has been handed to AI. It is one in which customer commitments remain connected to engineering and operational reality as circumstances change, without forcing manufacturers to dismantle the systems that already run the business.

The orchestration gap: Building the agentic manufacturing enterprise without replacing the core explores that operating model in greater depth, including quote-to-cash, managing by exception, governance and the route from individual use cases toward wider orchestration.

Download the new joint white paper from The Manufacturing Tech Brief and Infosys Topaz AI Next to explore how manufacturers can begin closing the gap between connected systems and connected execution.

Fill out this form to access the whitepaper