Jakub Zavrel, Founder and CEO, Zeta Alpha
The real question is not whether AI will have an impact, but where it can create the most value, and how to design it for industrial complexity, control, and sovereignty, and make the whole company and its people smarter.
Industry is currently under pressure from many sides at once: rising complexity, global competition, geopolitical tensions, and demographic change. Sometimes, it feels like the urgency to adopt AI is just yet another pressure point, but for most manufacturing companies, the real question is not whether AI will have an impact, but where it can create the most value beyond generic copilots and isolated proof of concepts.
For years, manufacturing companies have accumulated some of the richest and most defensible knowledge assets in the world: enormous amounts of valuable information across documents, technical manuals, engineering drawings, CAD/CAM, maintenance logs, ERP systems, PLM platforms, Sharepoint folders, and the tacit experience of long-tenured experts. In practice, much of this knowledge remains fragmented, inaccessible, and locked inside systems and teams. The result is a familiar problem: if only the company knew what the company knows. AI is fundamentally changing that equation now.
In AI for industry, the goal is not to help one employee write faster emails or summarize a meeting. The objective is to make the organization itself more intelligent: to create a secure, company-wide knowledge layer: a company brain, that connects data sources, understands technical context, and helps all employees find, use, and act on the right information; an AI that augments the intelligence of the company as a whole, and allows it to move faster and make better decisions.
At Zeta Alpha, this is exactly the area we focus on, by building the AI foundation layer for knowledge in industry, deployed at leading industrial companies such as Festo, BASF, OEDIV, Sartorius and Envalior. We are convinced the next wave of industrial AI will not be won by generic AI. It will be won by platforms that can integrate and customize deeply into enterprise environments, understand domain-specific knowledge, and deliver trustworthy answers and actions across the organization.
The bottleneck in industrial AI is usually not the AI model itself. In our experience, 80 percent of the challenge is data infrastructure: connecting fragmented systems, cleaning and structuring information, handling permissions correctly, and making knowledge accurately retrievable in context. Large language models from frontier labs like OpenAI, Anthropic, and their open source equivalents, are becoming more capable and more commoditized every month. The real differentiator is the application layer: how AI is grounded in your own enterprise knowledge, how well it is customized to your own industrial workflows, and how securely it can be deployed in the IT environments that you control. This is also prominently a question of sovereignty.
AI for industrial knowledge needs carefully tuned high-quality search, customizing advanced approaches such as agentic retrieval-augmented generation and task-specific deep research agent workflows. And it needs agents that do more than answer questions: investigate across multiple systems, synthesize findings, and support complex and long-horizon tasks.
Across European manufacturing, many experienced engineers, technicians, and plant specialists are approaching retirement. Decades of practical know-how often do not live in systems, but in people’s heads. At the same time, companies struggle to hire enough skilled experts, while products, regulations, and production environments become more complex. New employees need to become productive faster, but the knowledge they need is scattered across tools, teams, and repositories.
This is why AI for knowledge management should not be seen only as a technology question. It is also a collaboration and productivity problem at the organizational level. Better AI-mediated access to knowledge reduces dependency on individual gatekeepers, shortens response times, and helps teams work across functions more effectively. It can support onboarding, improve service quality, reduce repeated mistakes, and make expertise available where and when it is needed. In that sense, AI becomes a tool for strengthening and augmenting the human system of the company: helping experienced employees scale their knowledge, helping newer employees ramp up faster, and helping the organization operate with greater continuity despite workforce constraints.
Recent developments in AI agents are especially relevant here. In industrial settings, a useful agent is a system that can reliably navigate company knowledge, follow permissions, reason across sources, and help employees complete real work. For example, an agent might trace the root cause of a recurring issue by combining service reports, product documentation, engineering changes, and supplier information. Or it might support an engineer by gathering all relevant design, compliance, and maintenance knowledge before a decision is made. This is where AI starts to become institutional memory rather than just a chatbot.
A particularly important aspect of industrial expertise is engineering data. Much of the most valuable knowledge is not stored in plain text. It lives in CAD files, technical drawings, photos, diagrams, and 3D models. If AI is to become truly useful in manufacturing, it must be able to understand these modalities as well. This is why we are investing in making AI understand multimodal engineering data and even 3D models: because the future of industrial knowledge management depends on connecting language-based AI with the physical and technical reality of products, machines, and production systems.
And what about AI sovereignty? Many companies want to use AI, but they are rightly concerned about data security, compliance, intellectual property, and strategic dependence. In industrial environments, these concerns are not secondary; they are foundational. Sensitive product data, process knowledge, and customer information cannot simply be exposed to uncontrolled external systems - the so-called shadow IT of personal devices and AI subscriptions. Sovereign AI therefore means more than hosting models in Europe. It means giving companies control over where data flows, how systems are integrated and customized, which models are used, and how security and governance are enforced.
This also shapes the build-versus-buy question. Some companies assume they must either adopt a fully standardized tool that does not fit their needs, or build everything themselves. In reality, there is a third path: customizable secure platforms like Zeta Alpha that can be deployed on-prem and adapted to the company’s systems, workflows, and governance requirements without forcing every industrial firm to become an independent AI software developer.
The broader strategic point is this: Europe should not approach AI only as a user of generic tools built elsewhere. We should also build the application layer that reflects the needs of our industrial base. Raw models may become commodities. But the systems that make AI useful, secure, and deeply embedded in industrial work will create lasting value. This is where a new generation of strong European technology companies can emerge.
At HANNOVER MESSE 2026, this conversation is especially timely. The future of industrial AI will not be defined by hype, but by whether companies can make their own knowledge usable at scale. Those who succeed will not just deploy AI, they will redefine their industry by building the institutional foundation layer for the AI era of the industrial age.
The shift from fragmented silos to a unified Institutional Memory is no longer a vision, it is a competitive necessity being deployed today. See it for yourself at Hannover Messe: join our keynote on “How Frontier AI Models can Understand 3D CAD Models for Parts Reuse”, attend our panel on Sovereign AI for Industry, and witness these systems in action with a live demo at our booth (Hall 11, Stand E67).
Photo: Jakub Zavrel, Founder and CEO, Zeta Alpha
Bio: Jakub Zavrel is the Founder and CEO of Zeta Alpha, an emerging European AI native Enterprise Search and Knowledge Navigation platform that enables a fast growing customer base in High-Tech Industrial Manufacturing, Life Sciences, and Chemistry to securely deploy customized AI Agents on their internal knowledge. Jakub is an experienced AI researcher, entrepreneur, and technologist, and believes in building strong, independent, and long-term sustainable European technology companies and products.
