How can connected machines be transformed into truly intelligent manufacturing? In this eco interview, Igor Mikulina, CEO and founder of MicroStep Europa and President of the Board of Trustees of the IndustryFusion Foundation, discusses open standards, edge AI and the evolution of the smart factory towards increasingly autonomous systems.
Industry 4.0 has been discussed for many years, yet in practice numerous machines and systems still operate largely in isolation, particularly at SMEs. What are the biggest obstacles on the path towards truly connected manufacturing today?
The biggest obstacle today is no longer basic technical connectivity, but the lack of shared meaning in data. Machines from different manufacturers, years of manufacture and technology backgrounds provide data in different formats and structures. Other challenges include proprietary interfaces, the lack of unique identities for machines and components, and high integration costs.
For SMEs in particular, digitalisation must not become a bespoke IT project lasting several years. What is needed are scalable solutions that integrate existing machines, describe information semantically and enable data to be used in a controlled manner. Only when machines, materials, orders and processes are understood in a shared context can connected systems develop into a genuinely networked production environment.
A smart factory depends on machines from different manufacturers being able to communicate with one another. What role do open standards and cross-vendor solutions play, and why is interoperability so crucial to the digitalisation of industry?
Open standards are essential if companies are to shape their digitalisation independently of individual manufacturers and platforms. This is about more than simply transferring data on a technical level. Systems must also clearly understand which asset is being referred to, what a data value means and the rules under which the information may be used.
We therefore need interoperability at several levels: interfaces, data models, identities, access rights and business processes. Semantic Process Data Twins can link machine data with its technical and operational context. Combined with open standards and sovereign data spaces, this creates a robust foundation on which machines, software solutions and companies can work together securely.
Interoperability therefore prevents the emergence of new digital silos and is becoming the economic foundation of scalable industrial platforms.
AI is currently ushering in a new stage of industrial digitalisation. Where do you see the greatest potential when AI is deployed directly on machines or at manufacturing sites, and which applications are already realistic today?
The greatest potential of edge AI lies in analysing production data directly where it is generated. This reduces response times, protects sensitive operational data and enables decisions to be made even when a permanent cloud connection is unavailable.
Applications such as condition monitoring, anomaly detection, predictive maintenance, quality control, process optimisation and intelligent assistance for service teams and operators are already realistic today. The next stage of development will combine these capabilities with semantic Process Data Twins. AI will then receive more than just readings: it will also understand which machine and process they come from and the technical context in which they belong.
This will allow it to derive explainable recommendations for action instead of merely reporting statistical anomalies.
At the eco discussion on IoT and AI, you spoke about edge AI in industry. Looking several years ahead, how will the smart factory change when machines are not only connected and automated, but are increasingly able to make decisions autonomously?
The smart factory is evolving from an automated manufacturing facility into a distributed, agentic decision-making system. In future, machines and industrial AI agents will not only assess their own condition, but will also decide collectively how best to fulfil an order.
They will simultaneously take into account factors such as material availability, machine condition, technological suitability, capacity, energy consumption, quality, delivery time, costs and logistics. If in-house manufacturing is not economically or technically feasible, the system can prepare for the work to be assigned to qualified partners in a controlled manner within a trusted manufacturing network. This turns “Make or Buy” into a dynamic “Make or Share”.
This requires clear decision-making boundaries, verifiable rules, secure digital identities and sovereignty for the companies involved over their data. The factory of the future will therefore not be controlled centrally by a single cloud. It will be based on the interplay of intelligent edge systems, sovereign cloud services and interoperable data spaces.
Humans will remain the responsible decision-makers, while AI analyses complex relationships, evaluates options and prepares operational decisions or executes them within defined boundaries.
