AI Is Moving Into the Industrial Workflow
Artificial intelligence is increasingly being positioned as an operational technology rather than a standalone digital initiative. In industrial environments, its value is becoming clearer when it connects existing software, control systems, production data and human decisions.
The challenge is not necessarily a lack of technology. Many manufacturers and logistics organisations already operate numerous applications, machines and data platforms. The larger problem is the manual work required to move information and decisions between them.
Recent findings cited by Ace Workflow indicate that 26% of documented enterprise work hours, and as much as 44% in finance, could potentially be recovered through targeted automation. While these figures are not specific to manufacturing, they highlight a broader issue that industrial engineers encounter regularly: process inefficiency often exists at the interfaces between systems.
The Real Bottleneck Is Often Between Systems
Industrial automation has traditionally focused on automating individual machines, production cells and control processes. Modern AI introduces a different layer of optimisation by addressing the decisions and information exchanges surrounding those systems.
A production line may already have PLCs, DCS platforms, MES software, historians, warehouse systems and enterprise applications. Each system can perform its intended function effectively while the overall workflow remains dependent on spreadsheets, manual approvals, repeated data entry or operator intervention.
From an engineering perspective, this is an important distinction. Automating another isolated task does not necessarily improve the complete process. Greater value can come from connecting existing automation infrastructure so that information moves between systems with less manual intervention.
AI as an Intelligence Layer for Automation
This is where AI can complement conventional industrial automation.
Traditional automation is generally designed around defined conditions, sequences and control logic. AI can add a decision-support layer capable of interpreting changing operational conditions, identifying patterns and prioritising exceptions.
Adie Taylor of Arvato describes this approach as using AI as an intelligence layer that allows automation to respond to real-world variability while enabling people to concentrate on exceptions.
That concept is particularly relevant to logistics and manufacturing. Automated systems are highly effective when operating conditions remain within their programmed boundaries. AI can potentially help interpret situations that are more difficult to represent through fixed rules alone.
The engineering objective should therefore not be to replace deterministic control with AI. A more practical architecture is to retain deterministic control for safety-critical and time-sensitive functions while using AI for analysis, prediction, optimisation and higher-level decision support.
Where Industrial AI Can Deliver Practical Value
Several applications are already emerging as realistic areas for industrial AI deployment.
Predictive maintenance can combine historical equipment data, vibration measurements, temperature values, alarms and operating conditions to identify patterns associated with developing faults.
Computer vision can support inspection, defect identification, inventory management and process monitoring where visual information is difficult to process manually at scale.
Intelligent robotics can allow automated equipment to respond more effectively to changing material flows, warehouse conditions and production requirements.
Real-time optimisation can use operational data to identify constraints and recommend adjustments to production, logistics or resource allocation.
The common factor is data. AI cannot compensate for poorly structured information, missing signals or uncontrolled data sources. Industrial organisations therefore need to treat instrumentation, connectivity, data quality and system integration as part of the AI implementation rather than as separate technical concerns.
Legacy Infrastructure Should Not Be Ignored
One of the more practical points emerging from the London summit agenda is the importance of integrating AI with existing infrastructure.
Industrial facilities rarely have the opportunity to replace every PLC, controller, sensor, SCADA system or DCS platform simply to introduce a new AI capability. Long equipment lifecycles mean that modern digital layers must coexist with systems that may have been installed many years earlier.
This creates an engineering requirement for controlled integration.
Protocol gateways, historians, APIs, edge computing, industrial networks and data platforms can provide interfaces between established automation systems and newer analytical applications. The architecture must also account for cybersecurity, latency, data ownership and system availability.
In my view, this is one of the most important practical considerations in industrial AI. The winning solution will often be the one that integrates with the installed automation base without unnecessarily disturbing proven control functions.
Human Expertise Remains Part of the Control Architecture
Industrial AI should not be evaluated solely by how many tasks it can automate.
Operators, maintenance engineers and process specialists possess contextual knowledge that is difficult to encode completely into software. They understand abnormal operating conditions, equipment behaviour and production constraints that may not appear clearly in historical datasets.
AI can reduce repetitive analysis and highlight anomalies, but human intervention remains important when decisions involve unusual conditions, safety considerations or conflicting operational objectives.
This suggests a model in which AI handles high-volume information processing while engineers retain authority over defined operational decisions. Such an approach can also make AI adoption easier because it introduces automation progressively rather than requiring an immediate transition to autonomous operation.
Governance Must Develop Alongside Deployment
Scaling AI across industrial operations also creates governance requirements.
Companies need to understand what data an AI system uses, how recommendations are generated, who is responsible for acting on them and what happens when the system produces an incorrect result.
For industrial environments, governance must extend beyond conventional IT considerations. Functional safety, cybersecurity, access control, change management, traceability and operational accountability can all become relevant depending on how closely AI interacts with production systems.
Schneider Electric's participation in the AI and Industrial Automation discussion reflects this broader challenge. AI strategy cannot remain isolated within a technology department when deployment affects energy management, automation systems and operational processes.
AI LIVE London Focuses on Practical Deployment
AI LIVE: The London Summit 2026 will take place on 20-21 October at Olympia London, bringing together more than 2,000 executives and featuring more than 50 expert speakers, four executive workshops and live technical demonstrations.
The AI and Industrial Automation fireside panel on 20 October is particularly relevant to manufacturers, logistics operators and engineering teams evaluating practical AI applications.
The broader event theme, Technology + Human Purpose, also reflects a useful direction for industrial AI discussions. The question is increasingly less about whether AI can perform a particular task and more about where it should be introduced, how it should interact with existing automation and what responsibilities should remain with people.
The Industrial AI Opportunity Is Integration, Not Replacement
The next stage of industrial automation is unlikely to be defined simply by adding AI to machines.
The more significant opportunity lies in connecting information, decisions and existing automation systems into a more responsive operational architecture. AI can analyse large volumes of industrial data, identify patterns and support decisions, while PLCs, DCS platforms and safety systems continue to perform deterministic control functions.
For industrial engineers, this means AI adoption should begin with clearly defined operational problems rather than technology demonstrations. Identify the manual handoffs, establish the required data, define the decision boundary and then determine where AI can add measurable value.
That approach makes industrial AI less about replacing existing automation and more about extending what established automation infrastructure can achieve.
