ChatGPT And GenAI Started It, Agentic AI Kept It Going, Now MCP Can Deliver On The Promise At Scale

The 2025 Verdantix global corporate survey highlights the structural barriers that continue to constrain industrial AI adoption. When asked about the most significant challenges facing analytics and AI initiatives, 70% of respondents indicated that their technology ecosystem does not currently support the use of AI – while 78% answered that poor-quality or incomplete data present a major impediment.

Industrial AI still depends on better stack integration

For many industrial and manufacturing firms, the central integration challenge is not whether AI tools are available, but how to connect multiple data sources, applications and operational systems in a way that lets AI work across the enterprise rather than inside isolated pilots.

Our global survey findings reinforce this theme: the primary obstacle to industrial AI success is the difficulty of deploying AI effectively within complex IT and OT environments. Fragmented architectures, legacy systems and inconsistent data models continue to limit organizations’ ability to translate technical capability into operational value. In practice, this means that AI performance is now tightly linked to tech stack alignment. If ERP, MES/MOM, SCADA, historians, QMS and asset management platforms aren’t connected cleanly, AI will struggle to deliver reliable, repeatable outputs.

This challenge was already evident during the first wave of industrial copilots in 2024, before intensifying in 2025 with the emergence of agentic AI. The sheer number of specialized agents required to automate complex industrial workflows, combined with siloed and unstructured data sources, has stalled many organizations’ attempts to deploy and leverage this next-generation technology. Just 4% of manufacturing firms are experimenting with or piloting AI agents, and only 1% are scaling the technology, according to research by McKinsey & Company; work is needed to allow organizations to create impact through scaling deployments.

Why MCP matters for industrial AI deployment

Model context protocol (MCP) is the technology now coming to the forefront of this issue. The open-source standard allows AI agents to connect to tools, systems and data sources through standardized interfaces, reducing the need for bespoke integrations – acting in a similar fashion to PROFINET in the automation world. That’s why MCP matters beyond AI hype. It gives organizations a more practical way to have AI bridge their wider tech stack, especially where multiple industrial systems need to share context, data and actions without relying on one-off integrations each time a new workflow is introduced.

This ease of integration massively improves an organization's ability to field agentic AI, enabling effective switching between LLMs, reducing hallucinations, and improving governance and data security. Major industrial technology vendors such as AWS, Databricks, HighByte, Microsoft and Siemens are putting significant resources behind expanding MCP use. For buyers, MCP is best understood as an orchestration layer for connected AI workflows. It doesn’t replace core industrial systems, but it can make it easier to align them and expose the right data and actions to AI agents in a more controlled way.

By lowering integration friction, MCP enables organizations to deploy contextualized data services that can be consumed by multiple agents simultaneously. This supports faster experimentation, easier scaling and more consistent performance management. In principle, MCP allows firms to move from isolated pilots to repeatable, enterprise-grade AI deployments capable of delivering measurable ROI. However, MCP alone does not resolve the underlying data foundation problem. This is an important distinction for industrial firms. MCP can simplify how systems connect, but it cannot fix poor source data, unclear ownership or inconsistent models across the stack. Organizations still need a coherent digital architecture if they want AI to work reliably at scale.

What industrial firms need to scale AI successfully

The effectiveness of agentic systems ultimately depends on the quality and availability of enterprise sources of truth, including ERP, MES/MOM, historians, SCADA and asset management platforms. If these systems remain poorly integrated or inconsistently modelled, MCP-enabled agents will simply automate fragmented decision-making. For industrial and manufacturing organizations, this makes stack integration a strategic prerequisite. The aim goes beyond system connectivity for its own sake: a cleaner path for AI to interpret context, move across workflows and support decisions that reflect the reality of plant and enterprise operations.

The window for ‘first mover’ advantage in industrial AI is closing rapidly. Early adopters are now shifting from experimentation to scaled deployment, embedding agentic capabilities into maintenance planning, production optimization, engineering change management and supply chain coordination. Organizations that fail to modernize their data architectures risk being structurally disadvantaged as competitors institutionalize AI-driven operating models.

Technology providers are increasingly aligning their portfolios to support this transition. MES/MOM vendors such as Aegis Software, Critical Manufacturing, SAP and Siemens are expanding capabilities for contextualization, orchestration and semantic modelling. These platforms are becoming critical intermediaries between operational systems and MCP-enabled agent layers. This is especially relevant for industrial organizations asking what they can use to align the tech stack more easily. In many cases, the answer will involve a combination of stronger MES/MOM, industrial data management and asset or operations platforms that can act as cleaner intermediaries between OT data sources and enterprise AI applications.

What MCP means for industrial transformation leaders

The onus is now on industrial and manufacturing organizations to invest in the foundational digital frameworks and technologies if they have previously been reluctant to do so. Having well-integrated enterprise tools across PLM, ERP, MES/MOM, QMS and more is no longer a luxury for forward-looking firms – it is an operational requirement. MCP can and will fuel agentic AI adoption at scale, and organizations should make sure they are not being left behind.

For leaders planning next steps, the practical takeaway is straightforward: start by identifying the systems that hold the most important operational context, then assess how well those systems are integrated today and where MCP or similar standards could reduce friction in exposing that context to AI. Firms that approach MCP as part of a wider integration strategy, rather than as a standalone AI feature, are more likely to see durable value.