Governing Intelligence in Manufacturing: Why ITIL (Version 5) Is the Framework Industry 5.0 Has Been Waiting For

# AI

# Service Management

Industry 4.0 gave us smart factories. Industry 5.0 asks a harder question: who is responsible for what the smart factory decides?

July 9, 2026

Andrea Haddad

Industry 4.0 gave us smart factories. Industry 5.0 asks a harder question: Who is responsible for what the smart factory decides? That question does not have a technical answer. It has a governance answer. And most manufacturing organizations are not ready for it.

I have spent years working at the intersection of IT and operational technology in highly regulated, safety-critical environments. What I have seen consistently is not a shortage of AI ambition. It is a governance gap that sits directly above the plant floor and below the boardroom, in the space where intelligent digital systems make recommendations that influence physical outcomes and human decisions.

OT standards like IEC 62443, IEC 61508, and the Purdue Model govern how machines behave. They are rigorous, necessary, and mature. It is also worth being honest about the Purdue Model specifically: the original architecture from the 1990s has evolved significantly in practice. What most serious OT security and architecture practitioners work with today is a modernized version that accounts for cloud connectivity, IIoT platforms, and edge computing in ways the original model never anticipated. I reference Purdue as a reference architecture, not as a static doctrine, because the plant floor of 2026 does not look like the plant floor that the model was designed for.

What none of these frameworks govern, in any version, is who owns an AI-driven recommendation after it leaves the model. How risk is accepted across IT, OT, and the business. How an intelligent system evolves after go-live when the world it was trained on has changed. How a human operator knows when to trust the output and when to override it.

That gap is not an engineering problem. It is a service and governance problem. And ITIL (Version 5) is built to close it. What happens when that gap is not closed? Consider a predictive maintenance model that was deployed successfully and initially delivered good results. Over time, operating conditions changed, but no one was responsible for reviewing the model’s performance. For eight months, the model continued making recommendations based on outdated assumptions. Operators followed its guidance, and maintenance teams trusted the schedule. Eventually, an equipment failure occurred that the model was not prepared to predict. The problem was not the technology itself. The problem was the lack of clear ownership and oversight. This was a governance failure, and one that could have been prevented with basic accountability and service management practices.

AI Is a Service, Not a Project

The most expensive mistake I see organizations make with AI in manufacturing is treating it as a project. A project has a start date, an end date, a go-live, and a ribbon-cutting. After that, it belongs to someone else. Except in most organizations, no one is sure who that someone else is.

AI in a manufacturing environment is not a project. It is a service. It has consumers, operators, engineers, quality managers, supply chain planners, whose decisions depend on its outputs. It has dependencies that must be managed: data quality, infrastructure reliability, model currency, cybersecurity controls. It has a lifecycle that does not end at deployment. It degrades. It drifts.

The plant changes around it, and it needs to change with the plant. ITIL (Version 5) provides exactly the operating model this requires. A named service owner. Defined performance objectives. Explicit risk acceptance. Change controls that protect OT environments while allowing the model to evolve. Continual improvement practices that treat drift and degradation as service management problems, not technology failures.

If it creates value and risk, it must be governed as a service. That principle is what makes ITIL (Version 5) the right framework for Industry 5.0 intelligence.

The Governance Stack Manufacturing Actually Needs

I want to be precise about what ITIL (Version 5) does and does not do in a manufacturing context because the framework is frequently misunderstood. The most important point is this: ITIL (Version 5) does not replace existing governance disciplines, nor does it attempt to govern areas already covered by specialized frameworks and standards. Instead, it provides a service management structure that helps organizations connect risk management, architecture governance, AI controls, safety engineering, and continual improvement to the digital services and business outcomes they support.

This connective role is often what manufacturing organizations are missing. Many organizations have adopted the right frameworks, standards, and controls, yet still struggle with fragmented accountability, unclear ownership, inconsistent lifecycle management, and disconnected decision-making across IT, OT, engineering, and business functions.ITIL (Version 5) does not run PLCs. It does not replace safety engineering, control system design, or OT-specific standards such as IEC 61508, IEC 62443, ISA-95, or the Purdue Model.

What it provides is something manufacturing organizations have historically struggled to establish: a consistent operating model for managing digital capabilities as services throughout their lifecycle, with clear ownership, accountability, performance objectives, and continual improvement.

Other frameworks address specific governance domains. COBIT aligns technology investments and decision-making with enterprise objectives and risk appetite. CISM and ISO 27001 provide information security governance and risk oversight. The NIST AI Risk Management Framework and ISO 42001 address AI governance, model behavior, transparency, and accountability. Architecture frameworks such as TOGAF guide enterprise architecture, while ISA-95 and the Purdue Model provide structure for industrial system integration and segmentation. Safety frameworks define the controls required when technology influences safety-critical operations.

Each framework addresses a different aspect of governance. ITIL (Version 5) complements these disciplines by providing the service management structure that enables organizations to operate, support, improve, and derive value from digital capabilities over time. Together, they help ensure that technology, risk, people, and business objectives remain aligned as manufacturing environments become increasingly intelligent and interconnected.

What Governed Intelligence Actually Looks Like

Governance is not documentation. I want to say that plainly because the word "governance" makes many practitioners reach for a policy template when what is needed is operational behavior. In a well-governed AI-enabled manufacturing environment, every AI service has a named owner who is accountable for its behavior and its outcomes. There are model cards that describe not just what the model does but where it fails, what its confidence boundaries are, and what a human operator should do when the output is uncertain. There are clear override rules. There are runbooks that people actually use at two in the morning when something degrades unexpectedly, not documents that live in a SharePoint folder that no one has opened since the go-live celebration.

Humans remain the decision-makers. This is the defining principle of Industry 5.0, and it is fully supported by ITIL (Version 5)'s emphasis on co-creation of value across stakeholders. AI functions as a cognitive layer that analyzes operational data, supports Digital Twin simulations, predicts equipment failures and resource usage, and provides recommendations that assist human decision-making in real time.

On the subject of digital twins and simulation platforms: a digital twin or physics-based simulation environment used for digital rehearsal of manufacturing processes and for informing production planning or capital investment decisions is not a peripheral visualization tool.

It is a digital service that influences real decisions with real consequences. When a simulation is used to plan a facility modification, to optimize an energy-intensive process, or to rehearse a safety scenario before a change goes live, the governance requirements are identical to any other service that influences physical outcomes. It needs an owner. It needs performance objectives. It needs change controls. Platforms in this category are increasingly embedded in how manufacturers plan and operate, and the governance conversation around them is years behind where it needs to be.

At Purdue Levels 2 and 3, decision-supported AI enhances operations without assuming control authority. The human stays in the loop. The governance structure ensures the loop is never broken.

The Infrastructure Argument

AI-driven manufacturing places demands on infrastructure that most organizations have not fully accounted for. This is where the governance conversation becomes an investment conversation, and where ITIL (Version 5) earns its place at the executive table. A hybrid infrastructure model is not optional in Industry 5.0. It is the reference architecture. Edge and on-premises systems handle real-time control, safety systems, and AI inference at the plant level. Cloud platforms provide the scalability required for model training, enterprise analytics, and simulation-intensive workloads. Success depends on integrating these environments through secure, resilient, and low-latency connectivity.

Container-based platforms such as Kubernetes are increasingly used to support AI inference, edge analytics, and IIoT microservices. Virtual machines remain essential for ERP, SAP, MES, and many legacy OT applications. Private 5G is emerging as a key enabler for mobile robotics, autonomous guided vehicles (AGVs), and collaborative robotics environments that require reliable low-latency communications. At the device level, Single Pair Ethernet (SPE) is gaining traction as a scalable and sustainable connectivity option for Level 0 and Level 1 industrial assets.

The challenge is not selecting individual technologies. It is determining how infrastructure decisions support business outcomes while balancing performance, resilience, security, compliance, and cost. Architecture disciplines define how systems are designed and where workloads are deployed. ITIL (Version 5) helps ensure those decisions are evaluated through the lens of service value, operational requirements, and continual improvement.

In Industry 5.0, infrastructure is no longer a supporting function operating behind the scenes. It is a strategic capability that directly influences the performance, reliability, and trustworthiness of intelligent manufacturing services.

The Real Competitive Advantage

There is a version of this conversation that focuses on which AI model is most accurate, which vendor has the best platform, and which organization has the most impressive pilot. That conversation misses the point entirely.

The manufacturing organizations that will lead the next decade are not the ones with the most advanced AI. They are the ones with the strongest governance, the clearest accountability structures, and the deepest human trust in the systems they have built. Those qualities do not come from technology investment alone. They come from applying the right operating model with discipline and consistency over time.

ITIL (Version 5) is that operating model. Industry 5.0 is the context that makes it urgent. The challenge is no longer deploying intelligent systems. The challenge is governing them as they evolve, influence decisions, and become part of everyday operations. As manufacturing becomes more intelligent and interconnected, success will depend on more than technology. Organizations that thrive will be those with clear accountability, strong governance, and the trust to use intelligent systems responsibly at scale.