Uncategorized

AI governance doesn’t have to be overwhelming. Learn the five key pillars manufacturers need to confidently scale AI.

 

TAKEAWAYS:
● While manufacturers are confident in AI’s potential, many are still determining how to govern it effectively.
● AI should be built on the same foundations as good corporate governance, IT governance and data governance.
● Manufacturers that establish practical, fit‑for‑purpose AI governance will be better positioned to scale AI capabilities.

 

Governance for artificial intelligence tools and applications has become a practical requirement for any industrial company deploying AI at scale. The question for manufacturers is how to develop an AI governance framework that fits operational realities.

As more manufacturing companies implement AI across the factory floor and other parts of operations, governance is also increasingly a table-stakes issue. In the RSM Middle Market AI Survey 2025: U.S. and Canada, 87% of 134 manufacturing industry respondents said their organization currently uses generative AI tools (e.g., ChatGPT, Azure AI, Microsoft Copilot, custom solutions, etc.), either formally or informally, in its business practices.

Among respondents using generative AI tools, according to the data, the following percentages agreed or somewhat agreed that:

• Generative AI has impacted their organization more positively than expected (90%).
• They have the right staff in place to implement generative AI effectively (80%).
• They were prepared for compliance with emerging AI regulations (78%).
• They need outside help to get the most out of their generative AI solutions (75%).
• Generative AI has been harder to implement than expected (64%).

Twenty-eight percent of respondents said they had experienced negative or unexpected consequences implementing AI initiatives within their organization, and 24% of that group identified governance and risk challenges as one category of negative consequences.

Together, these findings suggest that while manufacturers are confident in AI’s potential, many are still determining how to govern it effectively as it moves from experimentation into core operations.

Effective AI governance in manufacturing starts with a simple premise: AI should be built on the same foundations as good corporate governance, IT governance and data governance.

 

But developing an AI governance framework—made up of policies, procedures and ethical controls that guide the responsible development, deployment and usage of AI systems—doesn’t need to be a daunting task. Effective AI governance in manufacturing starts with a simple premise: AI should be built on the same foundations as good corporate governance, IT governance and data governance. From there, manufacturers can focus on a set of core principles that apply across use cases, whether a model is predicting maintenance needs or optimizing energy usage.

Manufacturing Considerations

Many manufacturers are adopting AI faster than their existing governance structures can keep up. In practice, this often means models are built or purchased, connected to operational data, and deployed before there is clarity around ownership, accountability or ongoing oversight.

At the same time, the risk profile of AI in manufacturing is different from other industries. For example:

• AI models are increasingly connected to operational technology (OT) systems, sensors and equipment that do not always speak the same language.
• Manufacturers tend to run lean data and technology teams compared to financial services or technology companies, so what works for other businesses might not work in the manufacturing space.
• Data quality varies by plant, by line and sometimes by shift. When an AI model produces an unexpected result, there may be impacts on safety, quality or uptime.

Regulatory pressure is also increasing. While AI‑specific rules continue to evolve, manufacturers already operate under strict requirements related to data protection, cybersecurity, product quality and internal controls. AI does not sit outside those obligations. Governance is the mechanism that helps companies connect AI use cases to existing expectations around risk, compliance and controls.

Five Framework Pillars

A responsible AI governance framework typically addresses issues such as bias, consistency, data privacy, cybersecurity, clear ownership of AI models and outputs, regulatory compliance, and explainability while ensuring alignment with organizational values and ethics.

Those considerations are reflected across the following five key pillars of an effective AI governance framework:

Pillar 1: Data and AI governance policies and procedures

For manufacturers, data governance is often the make‑or‑break factor for AI. Operational data comes from many sources (sensors, machines, legacy systems, etc.) and it is not always consistent or well documented. Establishing traceability for operational data is critical. Even lightweight steps, such as data quality dashboards or basic documentation of data sources, can significantly reduce downstream risk. Interoperability is another practical concern. AI governance should account for how data moves across systems and processes, not just within a single model. Without that visibility, monitoring AI performance over time becomes far more difficult.

Pillar 2: AI oversight committee

Establish an AI oversight group with representation from operations, IT, data, security and the C-suite. This committee should be accountable for how AI is introduced, governed and monitored across the organization. The committee should regularly assess whether AI systems are delivering measurable operational or financial value, such as reduced downtime, improved yield or better forecasting accuracy.

Pillar 3: Third-party risk management enhancements

Many manufacturers will buy AI capabilities rather than build them. That makes third‑party risk management a core part of AI governance. Before deploying a vendor solution, manufacturers should understand what data the vendor can access, how that data is used and what commitments exist around bias management, accuracy and explainability. Service‑level agreements matter, but they only work if someone is responsible for monitoring them.

Pillar 4: AI system impact assessment criteria

Build risk and impact assessments into the intake process for new AI use cases so teams can identify potential issues before systems are deployed to the factory floor or enterprise processes. Assess whether the AI system influences safety‑related decisions, production continuity, quality outcomes or regulatory compliance. Higher‑impact use cases warrant stronger controls and monitoring.

Pillar 5: Stakeholder training and education

AI governance is not just for data scientists or executives. Operators, engineers, supervisors and anyone else who interacts with AI outputs need to understand what the system can and cannot do. Training should focus on practical issues such as how to interpret results, when to escalate concerns and how to recognize potential failure points. Teams need to lean into AI adoption just as much as they need to understand when to question the technology’s recommendations.

Preparing for the future

AI will continue to reshape factory operations, but its value will depend on how well companies govern it. Manufacturers that establish practical, fit‑for‑purpose AI governance now will be better positioned to scale AI capabilities responsibly and confidently as the technology matures.  M

 

About the authors:

 

Joseph Fontanazza is a manager at RSM US LLP.

 

 

 

Ciara Knight is a supervisor at RSM US LLP