The Growing Gap Between AI Investment and Business Results
Manufacturers are accelerating their investments in artificial intelligence. From appointing Chief AI Officers and establishing AI centers of excellence to deploying intelligent automation across production and maintenance operations, AI has become a major part of industrial digital transformation strategies.
However, greater investment does not automatically translate into measurable business value. McKinsey's latest research indicates that nearly two-thirds of organizations have yet to report a positive enterprise-level earnings impact from AI. This raises an important question: Is the challenge really the technology itself, or does it lie in how companies organize and manage AI-driven decisions?
In industrial environments, the distinction matters. An AI system may identify problems faster, process larger volumes of operational data, and generate recommendations more accurately. But these capabilities alone cannot guarantee better production performance or lower operating costs.
When AI Recommendations Conflict With Plant-Level Priorities
Consider a manufacturing facility where an AI-powered predictive maintenance system identifies an increasing probability of failure in a critical production machine.
The maintenance team recommends an immediate shutdown to prevent equipment damage. Production managers, however, are under pressure to fulfill customer orders and avoid schedule disruptions. The supply chain team reports that replacement parts will take three days to arrive, while finance is concerned about the cost implications of an unplanned stoppage.
Each department has a legitimate operational concern, but their priorities are not necessarily aligned.
An AI agent could evaluate equipment condition, production schedules, spare-parts availability, maintenance costs, and potential downtime. It might even calculate several possible scenarios and recommend the option with the lowest overall business impact.
Yet one fundamental issue remains: Who has the authority to approve that recommendation?
This is not simply a question of AI capability. It involves operational responsibility, decision-making authority, and the ability to balance competing business objectives.
From Predictive Maintenance to AI-Assisted Operational Decisions
Industrial automation has traditionally focused on collecting process data, monitoring equipment conditions, and maintaining stable production operations. PLCs, DCS platforms, SCADA systems, and condition monitoring solutions provide the information operators need to understand what is happening inside a plant.
Predictive analytics has extended these capabilities by identifying abnormal operating patterns and estimating potential equipment failures before they occur.
The next stage is more significant. AI agents are beginning to move beyond identifying problems and predicting outcomes toward evaluating alternatives and recommending operational actions.
For example, an industrial AI platform could combine vibration measurements, temperature trends, equipment operating history, maintenance records, and production requirements to assess whether a machine should continue operating or be taken offline.
However, there is an important distinction between generating a recommendation and executing a control action.
An AI-generated maintenance recommendation should not automatically become a machine shutdown command. Safety interlocks, protection systems, operating procedures, and authorized personnel must remain part of the decision process. The level of autonomy should be defined according to the application, risk assessment, and applicable safety requirements.
Organizational Accountability Is the Missing Link
One of the less discussed challenges in industrial AI implementation is the gap between technical capability and organizational responsibility.
IT departments may provide computing infrastructure and data platforms. Digital transformation teams may develop AI strategies, while automation engineers integrate AI applications with existing industrial systems. Maintenance and production teams contribute operational knowledge and evaluate recommendations.
But when an AI-generated decision affects equipment availability, product quality, production targets, or personnel safety, responsibility cannot be distributed so broadly that nobody is clearly accountable.
A practical industrial AI framework needs to establish several boundaries:
- Technical responsibility: Who develops, validates, monitors, and maintains the AI models and their data pipelines?
- Operational authority: Who approves changes to production schedules, maintenance plans, and equipment operating conditions?
- Safety responsibility: Which authorized personnel and protection systems have final authority over safety-related actions?
- Business accountability: Who evaluates whether AI implementation actually improves productivity, reduces costs, and supports long-term business objectives?
Without these boundaries, even a technically sophisticated AI solution can create confusion rather than improve operational efficiency.
The CEO's Role in Scaling Industrial AI
Chief Information Officers, Chief Digital Transformation Officers, and Chief AI Officers can lead technology development, establish governance frameworks, and coordinate AI deployment across the organization.
However, enterprise-wide AI transformation requires more than technical leadership.
The CEO must ensure that different departments work toward shared business objectives rather than optimizing their individual performance indicators. This may require changes to investment priorities, operational procedures, performance evaluation methods, and decision-making structures.
For manufacturers, this means evaluating AI not only by model accuracy or the number of completed projects, but also by measurable improvements in equipment availability, maintenance planning, production efficiency, energy consumption, and operational risk.
AI should support a coordinated operating strategy rather than become another isolated digital initiative.
An Industrial Engineer's Perspective: Keep Human Authority Connected to Machine Intelligence
From an industrial automation engineering perspective, the most important development is not simply that AI can analyze more data or generate more sophisticated recommendations. It is that AI is beginning to influence decisions that were traditionally made through collaboration between operators, maintenance engineers, production managers, and management teams.
This creates an opportunity to improve how industrial facilities respond to equipment failures, process deviations, and changing production requirements. However, it also makes system integration and governance more important.
Manufacturers should establish clear decision boundaries before expanding AI autonomy. AI can analyze equipment conditions, compare operational scenarios, and recommend corrective actions. Authorized personnel should retain control over decisions that require engineering judgment, particularly those involving process safety and equipment protection.
Over time, some low-risk and well-validated actions may be suitable for greater automation. Even then, traceability, access control, fallback procedures, and human oversight need to be considered during system design.
The real measure of industrial AI maturity will not be how many decisions an algorithm can make independently. It will be how effectively an organization combines machine intelligence with engineering expertise, operational authority, and measurable business accountability.
Ultimately, AI can help manufacturers determine what action makes sense. Deciding who is authorized to take that action remains a management responsibility.
