Factories are increasingly becoming software environments. Machines, production lines and maintenance teams already generate large amounts of data, but much of it remains fragmented or difficult to use in real time. A new generation of industrial AI companies is trying to change this by helping manufacturers understand what is happening on the shop floor, diagnose problems faster and support the workers responsible for keeping production running. This landscape looks at the European startups building these new intelligence and agent layers for factories.

Category #1: Production & Process Intelligence

What is this category about?

• The intelligence layer for factory operations. These companies help manufacturers understand what is happening on the shop floor and why production is not running as expected. The goal is to turn factory activity into decisions that improve how the plant operates.

• Making fragmented factory data useful. Production information is usually spread across machines, cameras and several software systems. These products bring that information together so factory teams can see how the different parts of the operation affect one another.

• Moving beyond traditional reporting. Most manufacturing software records what happened and presents it through dashboards. The new generation of products tries to explain the cause of a problem and suggest what the factory should do next.

• Improving the economics of production. The value usually comes from reducing downtime or increasing output. The same intelligence can also help a factory lower waste and react more quickly when something goes wrong.

• Addressing different types of manufacturing. Some companies focus on factories where workers assemble products manually. Others work with continuous processes such as chemical or energy production. In both cases, the product is designed to understand the production process and help it run better.

What do products in this category do?

• They usually complement the existing factory software stack. Manufacturers continue to run their operations through systems such as ERP, MES and SCADA. The intelligence platform connects to these systems and adds a new AI powered analytical layer on top of them.

• Some products use cameras to understand what happens on the line. Almetra applies computer vision to manual production processes that are often poorly represented in factory software. Cerrion uses video AI to detect process deviations before they turn into downtime or waste.

• Other platforms analyze production data to explain performance problems. CloEE helps factories understand downtime and production efficiency. Ethon uses causal AI to identify which variables are likely to be responsible for a loss in performance.

• Some companies are building a broader model of the factory. Quindi combines production information with data from systems such as ERP and WMS. Forgis brings together machine data and visual information so the platform can understand the factory as one connected environment.

• Planning platforms focus on what production should do next. Akeno creates schedules for complex manufacturing operations and updates them as conditions change. Zentio is developing a similar approach through an agentic model of the factory that can react to disruptions.

• A few products are beginning to act on their own recommendations. Most systems still keep a human in control. However, Akeno can already update production plans automatically, while Cerrion has connected its video intelligence to industrial control systems that can trigger corrective actions.

• Industrial foundation models are emerging in more specialized environments. Applied Computing is developing Orbital for refineries and petrochemical facilities. The model combines operational data with the physical rules that govern the industrial process.

• Deployment is part of the product. A factory may operate machines that are decades old alongside newer software systems. Vendors therefore need to support legacy infrastructure and offer deployment models that meet strict security requirements.

• Most customers start with one narrow use case. A first project may focus on one production line or one planning workflow. The vendor can then expand to other parts of the factory once the product has demonstrated measurable value.

What types of customers are using these products?

• Large manufacturers with manual or semi-automated assembly lines. ABB uses Almetra in electronics manufacturing, while Grundfos deployed it across a manual pump assembly line. These customers already collect production data but need better visibility into the human activity and small process losses that traditional factory systems miss.

• Process manufacturers with difficult planning constraints. Akeno works with companies such as BASF Coatings and Sun Chemical, where production planners must continuously account for changing demand and operational restrictions. Its broader target market includes chemical, pharmaceutical and food production.

• High-throughput plants where small disruptions quickly create waste. Stölzle Lausitz expanded Cerrion across four glass production lines after an initial pilot. Video agents are particularly relevant in this type of environment because a blockage or fallen product can rapidly affect safety and production output.

• Energy and chemical operators running highly complex processes. Applied Computing targets facilities such as refineries and petrochemical plants. Its partnership with KBR extends the product into ammonia production, where operators need to improve plant performance without compromising reliability.

What are the major trends shaping this category?

• European manufacturers are moving beyond the first wave of AI pilots. Many companies have experimented with AI, but relatively few use it deeply inside production operations. The market will increasingly be shaped by vendors that can turn an initial pilot into a system used across several factories.

• The pressure to improve productivity is becoming stronger. European factories face structurally higher operating costs than many international competitors. This makes software that improves the performance of existing plants increasingly valuable, especially when building new capacity is expensive.

• Connecting to existing factories remains the main technical challenge. Most European manufacturing sites are brownfield environments. Their machinery and software were installed at different times and were not designed to share data with modern AI systems.

• Production software is moving from observation toward decision making. Earlier tools mainly showed managers what had happened. New products are expected to explain the problem and help choose the response. Full autonomy will remain limited because mistakes can damage equipment or create safety risks.

• Digital twins are becoming operational rather than purely descriptive. The first generation recreated a digital representation of a factory. The emerging generation uses this model to simulate decisions and adapt production plans as conditions change.

• The loss of industrial expertise is increasing demand for these systems. Many European manufacturers rely on experienced workers who understand how machines behave in practice. As this workforce ages, companies need better ways to preserve that knowledge and make it available to less experienced employees.

• European data regulation may create opportunities for independent software vendors. The EU Data Act gives companies stronger rights to access data generated by connected equipment. Over time, this could make it easier for factory intelligence startups to build products on top of machines supplied by established industrial vendors.

• Europe remains a fragmented commercial market. Manufacturing strengths differ significantly between countries, as do local industrial networks and procurement practices. Startups will often need to build their go-to-market strategy around specific regions or industries before expanding across the continent.

How does the funding environment look?

• The category has attracted meaningful funding at several stages. Every company in this part of the landscape has raised external capital. The rounds range from relatively small pre-seed financings to larger Series A investments.

• The largest rounds have gone to companies with broad platform ambitions. Almetra raised a €16.3 million Series A in June 2026 after previously raising €4.5 million. Cerrion raised an $18 million Series A in November 2025 following a $5 million seed round.

• Applied Computing has recently gained strong momentum. The company raised £9 million in 2025 and another $20 million in July 2026. The latest round will support the international deployment of its industrial AI platform.

• Several companies are still developing the market with much smaller amounts of capital. Forgis raised a $4.5 million pre-seed round, while Akeno secured a €4.5 million seed. Zentio, CloEE and Quindi remain earlier and have raised more modest rounds.

• Investors appear most interested in products that can expand across factory workflows. A narrow analytical feature may be a useful starting point, but larger rounds have generally gone to companies that can become a broader operational layer.

• A meaningful share of the capital is being spent on deployment and commercialization. Looking at the hiring patterns of these companies, it seems that these startups are hiring people who can install the product and work directly with industrial customers.

• Consolidation has not started yet. I have not recorded meaningful acquisition activity among the companies on the landscape. Partnerships with established industrial groups are currently a more common route to distribution.

Category #3: Frontline & Maintenance Agents

What is this category about?

• The AI powered support layer for the people who keep factories running. These products help operators and maintenance teams understand equipment problems and decide what to do next. They bring relevant knowledge directly into the worker’s daily environment.

• Reducing the distance between a problem and its resolution. When a machine stops working, the answer may be hidden inside a manual or known only by one experienced technician. Frontline agents make that information easier to find and apply.

• Preserving industrial knowledge before it disappears. Much of the expertise inside a factory has never been formally documented. These products capture what experienced workers know and make it available to the wider team.

• Making maintenance less dependent on individual experts. The goal is not necessarily to replace the technician. It is to help more workers diagnose common problems without waiting for the most experienced person to become available.

• Supporting work beyond the factory itself. Some products are used by internal maintenance teams. Others help machinery manufacturers support customers after a machine has been sold. This extends the category into field service and industrial after sales operations.

What do products in this category do?

• They generally work alongside the systems manufacturers already use. Maintenance records may remain inside a CMMS, while technical documents stay in existing company repositories. The agent connects this information and gives the worker a simpler way to use it.

• Knowledge agents provide answers in the context of a specific machine. Acolyt and Edmund help technicians diagnose issues and follow the appropriate procedure. Their answers are connected to the equipment and operating environment rather than coming from a general-purpose chatbot.

• Some platforms focus on manufacturing service teams. lytra retrieves information from company documents and previous service cases. It can then help route a request or prepare a response for the employee handling it.

• Digital work instruction platforms are adding an AI layer. Flow Tool helps companies create procedures and deliver guidance while work is being performed. Picomto covers a similar workflow and can now turn a filmed intervention into a structured guide.

• Linexa focuses on the automation systems behind the production line. Its product reads PLC and SCADA project files so engineers can understand how the equipment has been programmed. The same environment can be used to investigate downtime or simulate a change before it reaches production.

• Rotomate begins with condition monitoring data. Rather than producing another alert, the product tries to explain the likely problem and recommend a maintenance response. This places it between traditional monitoring software and the technician who must decide what action to take.

• Synthavo applies AI to the spare parts workflow. A technician can photograph a component and identify the correct replacement without knowing its product number. The product then connects identification with the ordering process.

• Remberg combines the agent with the underlying maintenance system. It already manages assets and work orders as a CMMS. Its AI layer can now prepare maintenance tasks and turn technical documents into procedures.

• The product must be adapted to the customer’s operational context. Generic industrial knowledge is rarely enough. The platform needs to understand the customer’s machines and the way its teams already work.

• Most deployments still begin with a controlled pilot. A first project usually covers one site or one type of equipment. The vendor can expand once the customer sees that the product reduces resolution time or improves the quality of maintenance work.

What types of customers are using these products?

• Internal maintenance teams at large production sites. LIQUI MOLY uses remberg to move from reactive repairs toward planned maintenance, while ams OSRAM uses its AI Copilot to help technicians find machine knowledge faster. These customers often have documentation spread across several locations and depend heavily on experienced employees.

• Manufacturers operating similar equipment across several plants. Edmund works with packaging companies such as Amcor Flexibles and Model Obaly. The product is used to make technical knowledge available across production sites rather than leaving each factory to solve the same problems independently.

• Machinery manufacturers with an important after-sales business. HOLMER and MBO use Synthavo to help customers identify spare parts from a photograph. For these companies, the agent improves service after the machine has been sold and helps direct customers back toward the manufacturer’s parts business.

• Regulated manufacturers with distributed frontline teams. Daher uses Picomto to distribute current operating procedures across its aerospace sites, while Curium uses it for traceability and remote maintenance in radiopharmaceutical production. These customers need guidance that can be controlled and updated across sensitive operating environments.

What are the major trends shaping this category?

• The loss of industrial expertise is becoming a structural demand driver. Europe’s workforce is ageing while many manufacturers already struggle to recruit people with the required technical skills. Frontline agents offer a way to retain part of the knowledge that would otherwise leave the company when experienced workers retire.

• European industrial policy favors augmentation rather than full worker replacement. The European Commission’s Industry 5.0 vision places worker empowerment at the centre of production systems. This creates a natural role for agents that improve human decision making while leaving responsibility with the technician.

• Generative AI is changing how industrial knowledge is created. Earlier systems required someone to manually write and maintain every procedure. Multimodal models can now interpret technical documents or turn visual evidence from the shop floor into usable guidance. Research into manufacturing assistants also shows that model configuration and retrieval quality have a major effect on reliability.

• Predictive maintenance is moving closer to the actual maintenance workflow. Detecting a possible failure is only useful when the maintenance team understands the warning and knows how to respond. The market is therefore shifting toward systems that explain the diagnosis and connect it to a recommended action. Explainability becomes especially important when the equipment is safety-critical.

• The EU Data Act could open industrial equipment data to more independent providers. Since September 2025, European businesses have stronger rights to access data generated by connected machinery. They can also share this data with third-party maintenance providers. This could weaken the control that equipment manufacturers traditionally hold over the service data generated by their machines.

• Large industrial software vendors are entering the same market. Siemens is expanding its Industrial Copilot into production and maintenance workflows. This gives established vendors a distribution advantage and means startups will need to provide deeper operational value than a conversational interface alone.

• Brownfield integration will remain the main limit on adoption. Many European factories still rely on equipment that was never designed to feed data into an AI application. A good model has little value when it cannot access reliable machine context or connect to the customer’s maintenance process.

• Frontline adoption will depend on trust as much as technical accuracy. Workers need to understand where a recommendation came from and must be able to correct it when it is wrong. Products that are introduced without involving the technicians themselves risk becoming another software layer that the factory does not use. Europe’s humanncentric approach to industrial AI makes this particularly important for the category.

How does the funding environment look?

• The category remains relatively early. The Axomap timeline currently records approximately €43.6 million in disclosed funding across the nine companies. Only six of them have a financing round recorded.

• remberg accounts for most of the capital raised. The company secured a €2 million seed round in 2019 and an €11 million Series A in 2022. It added another €15 million in 2025, bringing its recorded total to €28 million.

• The other funded companies remain at seed or pre-seed stage. Augmented Industries raised €4.5 million, while Synthavo secured €4 million. Edmund has raised €3 million across two rounds.

• The newest companies are still financing initial product development. Rotomate raised a €2.1 million pre-seed round in June 2026. Linexa secured €2 million shortly before launching its shop floor intelligence platform.

• The category has not started consolidating yet. No acquisition involving these companies is currently recorded. Partnerships with industrial platforms appear to be the more immediate path to distribution, as illustrated by Flow Tool’s presence in the Siemens Xcelerator ecosystem.