Reshaping industrial operations
Across every sector, AI software is influencing how assets are maintained, how processes are optimised, and how workforce planning is transformed, but achieving these benefits is a different matter
The market is saturated with AI solutions promising value, but they may only work in isolation. Without cohesion, pilot projects often do not translate well into industrial reality, and so operators need a clear view of how AI is maturing to overcome this, as well as what defines a scalable project, and how to align multiple initiatives to unlock the benefits.
For example, there are new dark factories in China that are fully automated systems, optimising themselves without human input. This approach was previously only viable for small-scale pilot operations, but the development demonstrates how industrial AI can be applied to enterprises in the UK and Ireland seeking to improve efficiency across unrelated systems.
Scalable solutions
China’s dark factories were built from the ground up, under controlled conditions and designed around a specific manufacturing model. Most engineering and industrial leaders are working with brownfield and legacy sites, meaning the success of next-generation industrial AI will be defined by their readiness to scale.
Businesses need to build the necessary foundation before deploying AI, as without it, pilot projects will end at the initial scope and fail to deliver further value or integrate into other operations. AI readiness is its own process that requires:
- Structured data: AI depends on consistent and contextualised information. Context is vital, as it reflects how processes operate together rather than in single applications.
- Ownership of systems: Enabling defined accountability across IT and OT, ensuring AI insights are acted on by the right people at the right time.
- Alignment between digital strategies and OT realities: AI initiatives must be grounded in operational constraints to ensure insights are relevant to operators' day-to-day needs.
- Strong Guardrails: To prevent AI hallucinations from impacting critical production systems.
With this foundation, AI can operate across fragmented environments, accounting for legacy systems working alongside new assets, data and workforce silos, and complex supply chains. Achieving this requires confidence that models can be refined to meet changing needs. Next-generation solutions will expand on small, discrete pilot projects and integrate with the strategy to be applied enterprise-wide.
Semi-autonomous AI operations
Fully autonomous operations are only possible under specific conditions, whereas semi-autonomous operations can be achieved using the building blocks companies have developed through modernisation projects in recent years.
In semi-autonomous operations, an AI agent will be deployed to monitor conditions, identify optimisation, and recommend actions in real time, enabling companies to see the benefits of AI without a complete overhaul.
The distinction between semi-autonomous AI and a traditional pilot project is that the AI agent is not solving a specific problem in a controlled environment; instead, it can trigger maintenance activities, balance energy loads, or adjust controls, while human operators can schedule higher-value tasks. The AI agent also collects and analyses data to refine recommendations, creating a feedback loop that makes it more accurate over time.
Unified data environment
For AI to be effective, there are new considerations, as its success depends on quality, continuity, and contextually relevant streams of data.
Traditional analytics depended on historical datasets to identify trends and generate reports. This shift with AI involves interpreting live conditions as they evolve. Many organisations face the challenge of a lack of accessibility and contextualisation, rather than a lack of data. AI requires a unified data environment where data from previously unconnected silos is collected, stored, structured, and accessed. This enables the AI algorithm to connect previously unconnected inputs and outputs.
Investing in data management will put a business in a better position to scale, as its AI model can use the same structure. New lines, machines, or components can be integrated into the unified data model without rebuilding it.
The unified data environment enables multiple AI solutions to work together. Successful AI scalability is about cohesion; businesses can see better results by creating a unified data environment to underpin AI.
All-important human operators
AI can evolve into a collaborative team member, working in tandem with human expertise to deliver the best possible results. The same can be said of China’s dark factories: even the most advanced AI and robotics can only deliver value because of the professionals who design systems, define parameters, and monitor and interpret progress. AI can analyse vast volumes of data, but human expertise is needed to apply insights and make judgements that affect production.
As industrial companies adopt AI, it is important to evolve and update operators’ roles and responsibilities as the technology advances. People need to provide feedback, intervention, and guidance to improve AI solutions. When hiring and training operators, they should have confidence in using and validating AI’s recommendations. The future of industrial AI will be driven by people and AI agents working together.
When solutions are introduced with the right digital foundation in place, operators can trust the algorithm’s outputs and that they reflect the whole business. Building confidence in AI is technical as well as human, and it relies on alignment between capabilities and industrial realities.
Deploying AI at scale
As solutions continue to evolve, the challenge for UK businesses is not access to AI but readiness for it. Successful businesses will focus on getting the foundations right: ensuring high-quality data in the right context and making it easily accessible, deploying cyber-resilient architectures, and building operational alignment before scaling AI initiatives. In turn, these companies can expand on isolated pilots to realise the full potential of industrial AI, creating an environment that brings together data and tools into a cohesive, intelligent system.
Contributed by Charlotte Smith, technical manager, SolutionsPT.
