Warren Blanchard
National Manufacturing Practice Co-Leader
Manufacturers are investing more in robotics, automation, and artificial intelligence (AI) to close capacity gaps, enhance consistency, and minimize hazardous work. But as our report, Managing Risk in Manufacturing, explains, AI and automation bring a range of risks that should be recognized, evaluated, and addressed. Here we highlight the main AI and automation risks manufacturing leaders should know and practical steps to consider. For a deeper analysis, case studies, and helpful checklists, download the full report.
Understanding the risks of AI and automation in manufacturing
How is this technology being used?
AI and automation are increasingly integrated into the shop floor, supply chain, quality control, and maintenance processes. Common applications include predictive maintenance, demand forecasting, process optimization, robotics, and optical inspection systems that replace manual visual checks.
Why does this technology amplify risk?
When AI or automated systems fail, consequences include increased downtime, safety incidents, loss of intellectual property (IP), regulatory or compliance lapses, and reputational harm. Separately, automation that replaces tasks without accompanying retraining can harm workforce morale and damage a manufacturer’s reputation as an employer.
What is the business impact?
Failures in AI or automated systems can lead to production delays and revenue loss due to extended downtime and quality issues. Safety incidents and regulatory lapses can result in operational and legal costs, while breaches of sensitive data or IP might impact your competitive edge and finances over time. It’s also important to consider the workforce: without clear reskilling pathways, automation can affect morale and make it harder to recruit and retain talent, which can impact productivity and your reputation as an employer.
Data risks and model reliability
AI and automation performance depend heavily on the quality of the data used for training. Inaccurate, biased, or incomplete datasets can lead to misclassifications, scheduling issues, and unsafe outputs.
Models can degrade over time. Model drift happens when data distributions shift or when training data no longer reflects current conditions. Without regular monitoring, these changes might go unnoticed for a while, so it’s important for teams to implement continuous validation and set up alerts to identify and address drift.
To effectively manage model risk, manufacturers can adopt validation and testing protocols and view models as part of a lifecycle that requires regular retraining and oversight. The report suggests processes for model validation, tracking data origins, and monitoring to ensure systems remain reliable in production.
Security, privacy, and IP protection
Cybersecurity threats related to AI and the convergence of operational technology (OT) and information technology (IT) are serious. Connected robots, PLCs, CAD/CAM systems, and cloud analytics platforms create multiple entry points for potential threats. Ransomware and supply‑chain issues that once affected office systems can now disrupt production lines, compromise batches, or expose production networks shared with vendors.
Protecting sensitive data and IP is crucial when training and deploying models. Model development often involves operational data and proprietary information, so manufacturers can benefit from implementing safeguards for training datasets and ensuring that IP protections are clear in vendor relationships and cloud platforms.
Basic security measures include access control, encryption, and incident response planning. It’s important to establish access controls and encryption, deploy models securely, and incorporate OT scenarios into incident‑response plans and tabletop exercises. Additionally, make sure that contracts and service level agreements (SLAs) reflect security and maintainability expectations.
Operational risks and safety
Real-time control systems and robotics present unique challenges in reliability and robustness. When machines fail, errors can escalate more quickly than human mistakes, potentially leading to the production of defective products or unsafe operating conditions before the problem is identified.
As people and autonomous machines work together, new types of accidents may arise. Standards like ISO 10218 and ISO/TS 15066 offer guidance for integration, but successful real-world deployments still require thoughtful engineering, active operator oversight, and customized safety analyses.
Manufacturers can effectively manage change by implementing controlled testing, establishing clear rollback and containment plans, and maintaining manual fallbacks for critical production paths. Conducting recovery drills and evaluating how operations could adapt if technology fails are practical steps to strengthen resilience.
Governance, compliance, and workforce impacts
Creating effective AI and automation governance involves a collaborative framework that includes operations, safety, IT/OT, legal, and HR teams. It’s important to categorize automation projects by their risk levels so oversight matches potential exposure. Additionally, defining clear roles, risk-acceptance criteria, and audit trails for AI-related decisions is essential.
Compliance with regulatory, privacy, and industry-specific standards should be closely linked to how models are trained and how data is managed. Manufacturers can benefit from aligning model development, data practices, and vendor contracts with relevant privacy rules and sector requirements.
As technology evolves, workforce training and reskilling become vital to support new roles such as robot technicians, PLC programmers, data analysts, and cybersecurity specialists. Pairing automation rollouts with clear, phased retraining and job redesign can help protect job quality, maintain morale, and reduce social and reputational risks.
Building AI and automation resilience with governance, workforce, and security
Bringing these elements together highlights that successful AI and automation adoption depends as much on governance, workforce strategy, and resilience as it does on technology. Practical steps—such as implementing model lifecycle controls, enhancing OT security, focusing on safety engineering, ensuring vendor diligence, and providing phased reskilling—transform abstract risks into a manageable program that protects operations and IP.
For a closer look at the case studies, implementation checklists, and insurance considerations that support this approach, download the full report or reach out to a Marsh McLennan Agency manufacturing specialist for a personalized risk assessment and mitigation plan.
