How AI Is Transforming Nordic Manufacturing: From Automation to Intelligent Production
At a factory in Hesselager, Denmark, C.C. Jensen faced a practical manufacturing challenge: knowing when its natural-fibre oil filters had dried sufficiently. Because the fibres did not dry at a uniform rate, using a fixed drying time could mean keeping the ovens running longer than necessary.
Working with Aarhus University through a MADE FAST research project, the company used sensors and a neural-network model to estimate when the filters had reached the right level of dryness. According to the project report, the company reduced energy consumption for the drying process by a factor of ten and cut average drying time per filter by 37%.
The results illustrate a practical application of artificial intelligence (AI) in manufacturing. Rather than replacing an entire production system, AI can help improve an existing process by making better use of operational data.
While AI in business is often associated with content creation and software development, its industrial applications are considerably broader. On the factory floor, AI models can analyse sensor readings, machine images and production records to support decisions about product quality, equipment maintenance, energy consumption and material utilisation.
Across the Nordic region, manufacturers are exploring these applications in different ways. The important question is not simply where AI is being used, but what it changes in the production process and whether those changes produce measurable benefits.
Beyond Automation: What AI Adds to Manufacturing
Automation has long been central to modern manufacturing. Industrial robots, conveyor systems and programmable controllers can perform repetitive tasks with speed and consistency. Many systems also use sensors and feedback mechanisms to respond to changing operating conditions.
AI adds another capability: learning patterns from data to support predictions, classifications and decisions that may be difficult to capture through fixed rules alone.
For example, a conventional inspection system might check a component against predefined criteria, while an AI-based vision system can analyse images to identify patterns associated with potential defects. Similarly, a predictive-maintenance model can use machine data to help operators recognise conditions that may indicate a developing problem.
AI and automation are therefore not competing approaches. They can work together, combining established industrial equipment with data-driven insights to improve production decisions.
The Nordic manufacturing examples below show how this combination is being applied to specific operational challenges.
Where AI Is Making a Difference
Four applications stand out in the available examples from Nordic manufacturers and industrial research projects.
| Application | What AI helps with | Example |
|---|---|---|
| Process optimisation | Estimating process conditions and avoiding unnecessary energy use | C.C. Jensen, Denmark |
| Quality inspection | Identifying defects and improving inspection consistency | Horse Powertrain, Sweden |
| Predictive maintenance | Monitoring equipment conditions and planning maintenance | Talgø MøreTre, Norway |
| Material utilisation and planning | Matching variable raw materials and production decisions to desired outputs | Metsä Fibre, Finland, and Völur, Norway |
These applications address different problems, but they share a common principle: using data to improve decisions that affect production quality, efficiency or resource use.
AI on the Nordic Factory Floor
Denmark: Optimising energy-intensive processes
C.C. Jensen's drying process provides one of the clearest examples of a measurable AI-related manufacturing improvement.
The company's industrial oil filters contain natural fibres that dry at different rates. Determining when the drying process was complete was important for product quality, but continuing to heat the filters after they had reached the required dryness wasted energy and production time.
Through its collaboration with Aarhus University, C.C. Jensen developed a neural-network model trained using drying experiments and sensor data. The model estimated when the filters had reached the appropriate level of dryness, helping operators avoid unnecessary heating.
The project report attributed a tenfold reduction in energy consumption for the drying process and a 37% reduction in average drying time per filter to the improvements achieved through the project.
These figures are specific to the drying process, rather than the factory's total energy consumption. They also describe the outcome of the broader project, so the entire improvement should not automatically be attributed to the AI model in isolation.
Nevertheless, the case demonstrates how AI can improve an established industrial process without requiring a complete production-system overhaul.
Source: Aarhus University — Collaboration reduces energy consumption by a factor of 10.
Finland: Using AI to optimise timber processing
At Metsä Fibre's Rauma sawmill in Finland, machine vision, X-ray imaging, AI, robotics and integrated control systems support timber processing.
The technologies help the facility assess incoming logs, select suitable end products and maintain quality throughout production. This is particularly relevant in timber processing, where raw materials naturally vary in their characteristics and the products that can be obtained from them.
The sawmill has a stated annual production capacity of 750,000 cubic metres of pine timber. That figure illustrates the scale of the operation, although it should not be interpreted as a capacity increase caused by AI.
The more interesting aspect is how data supports decisions about the material itself. By combining information from imaging and production systems, manufacturers can make more informed choices about how individual logs should be processed to obtain suitable end products.
This application extends AI's role beyond detecting defects. It shows how data-driven decisions can support material selection, product quality and production planning.
Sources: Metsä Fibre — Rauma brings machine vision to the sawmill industry; Metsä Fibre — Rauma sawmill.
Sweden: Improving inspection and predicting defects
Sweden provides examples of both AI systems used in manufacturing and research into how machine learning could help prevent defects.
AI-assisted visual inspection
Horse Powertrain's engine production facility in Skövde uses Siemens' Inspekto AI-based vision system to inspect components. The system checks components within seconds and provides inspection results that support more consistent and traceable quality control.
This demonstrates how AI-based visual inspection can become part of an existing production workflow. Rather than relying exclusively on manual checks, manufacturers can use image analysis to identify potential deviations more quickly.
The published case study describes the inspection capabilities but does not quantify a reduction in defect rates or overall production costs. The benefit should therefore be framed around inspection speed and consistency rather than an unverified financial outcome.
Source: Siemens — AI-powered vision system at Horse Powertrain.
Predicting manufacturing defects
Research at Chalmers University of Technology has examined how machine-learning models can identify production conditions associated with casting defects in heavy-vehicle powertrain components.
The study analysed production parameters and data from core-making machines to investigate whether patterns in the manufacturing process could help identify conditions linked to defective components.
The potential benefit is straightforward: identifying problematic conditions earlier could help manufacturers investigate the cause of defects before affected components move further through production.
However, this should be understood as a research application rather than evidence of a fully deployed, factory-wide defect-prediction system. Its significance lies in demonstrating how production data can be used to investigate quality problems that may not be immediately visible during inspection.
Source: Chalmers University of Technology — Manufacturing quality research.
Norway: Using machine data to support maintenance decisions
At Talgø MøreTre, a Norwegian lumber producer, equipment reliability became increasingly important as production volumes grew. A splitting saw was identified as a bottleneck because overheating could affect product quality and lead to unplanned downtime.
Replacing a saw blade and allowing the equipment to cool could result in approximately two hours of downtime. The company also faced the prospect of investing in a new splitting saw to address its production constraints.
A research study examined how a predictive-maintenance platform could help address this problem. The solution used sensors to monitor parameters including temperature, vibration and power consumption. The collected data was made available to operators, while temperature thresholds provided early warnings when the equipment approached potentially problematic operating conditions.
This gave operators an opportunity to adjust the machine's operating speed and plan maintenance before a failure occurred, rather than responding only after a problem developed.
The study reports that the company moved from a reactive maintenance approach towards more data-informed planning. The improved understanding of equipment conditions also helped the company increase production capacity to a level at which investment in a new machine could be postponed.
This is an important example of how predictive maintenance can support production decisions. It also illustrates that the practical value of industrial AI may come from helping operators use existing equipment more effectively, rather than replacing machinery or removing human involvement.
Source: Springer Nature — Application of sensor-data-based predictive maintenance and artificial neural networks.
Norway: Optimising production planning
AI can also support manufacturing decisions that involve complex combinations of raw materials, production capacity and changing customer demand.
Völur, a technology company based in Oslo, develops AI software for meat-processing operations. Meat producers must determine how to cut and allocate raw materials to produce the right products in response to market demand. Poor planning can leave producers with products that are difficult to sell while increasing waste or reducing the value obtained from available materials.
Völur's software uses data from individual processing facilities to generate daily sorting and cutting plans. These plans help match available raw materials with the products that are most suitable for current demand.
According to Innovation Norway, Völur reports that daily optimisation can increase operating revenue by 2–4%. This is a company-reported figure, not an independently verified result that should be assumed to apply to every customer.
The example demonstrates another role for AI in manufacturing: supporting decisions about how resources should be allocated across a complex production process. Rather than focusing only on machine performance, AI can help manufacturers plan how to obtain greater value from the materials they already have.
Völur's international customer base also illustrates why this case should be treated as an example of Norwegian-developed industrial technology, rather than evidence that all of its reported results come from Norwegian factories.
Source: Innovation Norway — AI puts the right product on the shop shelf, and less in the bin.
What Results Are Manufacturers Actually Seeing?
The examples show that AI can contribute to different manufacturing outcomes, but the available evidence is not equally strong in every case.
Energy and production time. C.C. Jensen's project provides the clearest quantified result, with reported reductions in energy consumption for drying and average drying time per filter. It demonstrates how better process estimates can help reduce unnecessary operation.
Quality inspection. Horse Powertrain's system performs component checks within seconds and provides traceable inspection results. This illustrates the potential for faster and more consistent inspection, although the published case does not quantify a reduction in defects.
Equipment reliability. The Talgø MøreTre study describes how sensor monitoring and early warnings helped operators plan maintenance and manage a production bottleneck. The reported benefit includes postponing investment in new equipment, rather than a quantified percentage reduction in downtime.
Material utilisation and planning. Metsä Fibre's production systems support log assessment and product selection, while Völur's software generates cutting and sorting plans based on demand. These examples demonstrate how data can support decisions about raw materials, although their outcomes are not directly comparable.
Taken together, the cases suggest that AI can improve specific manufacturing processes. They do not establish a single productivity gain that can be applied to every factory or production line.
The distinction matters because a system that improves inspection in one setting may deliver different results in another. The outcome depends on the problem being addressed, the quality of available data and how the technology is integrated into the production process.
Why Industrial AI Is Difficult to Scale
Demonstrating that an AI system works in a specific setting is only one step. Applying it reliably across an entire manufacturing operation introduces additional challenges.
A 2026 research review examining industrial AI for maintenance in Sweden highlights practical barriers to wider implementation, including data availability, implementation costs, specialist expertise and organisational readiness.
Several of these challenges are particularly relevant to manufacturing.
Data quality and availability
AI models rely on reliable information about the processes they are intended to support. Machine readings may be incomplete, historical maintenance records may be disconnected from production systems, and different equipment may store data in incompatible formats.
Without sufficient and reliable data, even a technically advanced model may produce results that are difficult to trust.
Integration with existing equipment
Manufacturing facilities often operate machinery installed over many years. Connecting modern AI systems to older equipment, sensors and production software can require substantial engineering work.
The challenge is not simply installing an AI model. It is making sure that the model receives the right information and that its output can be used effectively within the existing workflow.
Reliability under changing conditions
A model that performs well during a trial may behave differently when raw materials, machine conditions or production requirements change. Manufacturers need to validate performance under real operating conditions and monitor whether results remain reliable over time.
This is particularly important when incorrect predictions could affect product quality, equipment safety or production continuity.
Return on investment
A successful technical demonstration does not automatically justify a wider rollout. Manufacturers need to establish whether the system produces sufficient improvements in energy consumption, quality, downtime or production efficiency to justify the cost of implementation and maintenance.
These challenges help explain why industrial AI should not be measured only by whether a model can make a prediction. Its value depends on whether that prediction leads to a useful, reliable and economically justified decision.
Source: Chalmers University of Technology — Industrial AI for maintenance in Sweden.
Solving the Right Manufacturing Problems
The Nordic manufacturing examples show how AI is being applied to specific operational challenges, from reducing energy use in Denmark to supporting quality inspection in Sweden, timber processing in Finland and predictive maintenance in Norway.
The strongest results emerge when the technology addresses a clearly defined problem and its contribution can be evaluated against practical outcomes. In some cases, those outcomes are measurable, such as shorter drying times. In others, the evidence demonstrates a useful application without providing a comparable numerical improvement.
The next challenge is to establish where these results can be repeated reliably, how AI can be integrated into existing production systems, and whether the benefits justify the cost and complexity of wider implementation.
For manufacturers, the goal is not simply to introduce more AI. It is to use data and intelligent systems where they can make production safer, more efficient, more consistent or less wasteful.
