Polish factories want to invest in AI, yet only a few percent of companies use it. The problem is not the algorithm, but money, data and automation. What to do so that the first AI project on the shop floor makes sense.
Big plans, small scale
Polish industry says it is interested in artificial intelligence, but the numbers show a clear gap between plans and reality. According to the ABB Trends 2026/2027 report (an IQS survey of 146 industrial companies), 34% of companies name AI as an investment priority for the next three years, while only 6% use AI and machine learning today (via ITwiz).
Eurostat paints a similar picture: in 2025, 8.4% of Polish enterprises used AI, compared with an EU average of 20%. The gap between small companies (5.8%) and large ones (45.8%) is huge (via ITwiz).
From the perspective of someone who has been responsible for IT in manufacturing companies for more than 20 years, these numbers are not surprising. AI in a factory does not start with a language model. It starts with a sensor, a controller, and data that can be read reliably.
The foundation that is often missing
The same ABB report shows the starting point of Polish plants:
| Technology | Share of companies using it |
|---|---|
| SCADA/DCS systems | 49% |
| Remote monitoring and diagnostics | 45% |
| Predictive maintenance | 28% |
| Industrial Internet of Things (IIoT) | 18% |
| Digital twins | 7% |
| AI and machine learning | 6% |
Source: ABB Trends 2026/2027, via ITwiz.
One in five manufacturers rates its automation as minimal, and almost one in three as low. Without automation and systems that collect data from the line, there is nothing to analyse — whether with classic methods or with AI.
EY’s figures are even more telling: 77% of medium and large companies plan to increase AI spending within 18 months, but only 9% have a complete data infrastructure (via ITwiz).
The three real barriers
Money
A limited budget is the biggest obstacle according to ABB — half of the companies cite it. The European Investment Bank’s EIBIS survey also shows that Polish companies finance investment mainly from their own funds (74%, versus an EU average of 66%), and most capital spending still goes into buildings, machinery and equipment (via ITwiz). In that structure, an AI project has to compete with a new production line — and it usually loses unless it has a clear business case.
Data
According to Deloitte, 30% of companies report problems with data availability and quality (via ITwiz). In practice, this means data scattered across controllers, the MES, the ERP and spreadsheets, stored in different formats, at different frequencies and without common identifiers for machines or production batches.
People
A lack of skills is cited by 35% of companies in the Deloitte study, and the availability of skilled workers is an investment barrier for as many as 81% of companies in the EIBIS survey (via ITwiz). AI in manufacturing needs people who understand both the technology and the production process. There are few such people on the market.
Where AI already delivers value on the shop floor
The most mature use cases are not spectacular, but they deliver measurable results:
- failure prediction and predictive maintenance — servicing a machine before the line stops,
- anomaly detection in process parameters, energy consumption or product quality,
- process optimisation — machine settings, schedules and raw material usage,
- quality control with machine vision — detecting defects on the line.
Deloitte notes that 84% of manufacturers achieve measurable operational value from AI, but only about one in five deployments is scaled across the whole enterprise or multiple plants (via ITwiz). It is a familiar pattern: the pilot succeeds, then stalls.
How to start so you don’t end at the pilot
I agree with the point made by Marek Sulewski of ABB, quoted by ITwiz: the biggest challenge today is not AI itself, but making effective use of the data companies already have. In my experience, this sequence works best:
- Start with a problem, not a technology. A specific, measurable goal: fewer unplanned stoppages on one line, fewer quality complaints, lower energy use in one process.
- Check what data you already have. SCADA, MES and ERP often hold more information than people think — nobody has connected it yet.
- Fix the foundation. Common machine and batch identifiers, time synchronisation, one place to collect data. It is dull work, but without it no model will be reliable.
- Run a pilot with a success metric. Before the project starts, decide how you will know it worked — and how much it must deliver to be scaled to other lines.
- Plan for scale from day one. A solution that works on one machine thanks to manual tweaks will not transfer to the whole plant.
- Bring IT together with maintenance and process engineers. AI projects in factories succeed when the data people and the process people sit at the same table.
- Don’t forget security. Connecting machines to networks and the cloud opens new attack vectors. Many manufacturers are also subject to the new requirements of Poland’s KSC Act implementing NIS2 — I wrote about it in NIS2 for mid-sized companies: where to start.
Summary
Polish industry is not short of willingness to use AI. It is short of foundations: automation, data, skills and funding. The good news is that each of these foundations delivers value on its own — better data and automation improve production whether or not an AI model is ever built on top of them. Companies that put their data and processes in order now will be able to deploy AI faster and more cheaply than their competitors in two or three years.
Sources
- Polski przemysł chce AI, ale brakuje mu pieniędzy i danych — ITwiz, 8 October 2026 (in Polish) — data from the following reports is cited via this article: ABB Trends 2026/2027 (IQS, CAWI, 146 industrial companies), EY “How Polish companies implement AI”, Eurostat (2025), Deloitte, and EIBIS 2025 (European Investment Bank).
✦ This text was prepared by an AI agent and reviewed by the author before publication.