"How widespread is AI-powered process automation really, and what do companies use it for?" This question sits on the desks of many management teams planning their automation budgets. The answer can be read from official statistics rather than from vendor presentations. Eurostat's report "The Use of Artificial Intelligence (AI) Technologies in the European Union – 2026," published on March 26, 2026, measures process automation with AI-based software robots (RPA) as a separate category. According to the report, the share of EU enterprises using this technology rose from 4.2% in 2024 to 5.4% in 2025; among large enterprises, the share is 24.4%.
Momentum is clear in Türkiye as well. According to the bulletin that TÜİK (the Turkish Statistical Institute) published on October 2, 2026, the share of enterprises using any AI technology rose from 7.5% in 2025 to 14.0% in 2026. In this article, we read the European data specifically through the lens of RPA and propose a practical framework for choosing the first process.
First, the definition: what does Eurostat measure?
Eurostat's survey category is broader than a "screen bot" in the narrow sense. The report defines this technology as process automation with AI-based software robots, under the heading of technologies that automate different workflows or assist in decision-making. Because the category name emphasizes AI-based software, this figure should not be read as the total prevalence of classic rule-based bots that contain no AI; what is measured is automation with an AI component that runs a workflow or supports a decision.
The basis of the data matters too. The survey covers enterprises with 10 or more employees and was conducted among 157,000 enterprises across the EU in 2025. For this reason, the figures offer a far broader picture than individual case examples.
The data: adoption rises sharply with scale
RPA is not yet mainstream in Europe. In 2025, 20.0% of EU enterprises used at least one AI technology, and this share rose by 6.5 percentage points compared with 2024. By technology type, text analysis is the most common at 11.8%, while workflow automation and decision support sit in the middle of the ranking at 5.4%.
What stands out most is the difference by scale. In 2025, the usage rate of workflow automation was 4.1% among small enterprises, 4.8% among medium-sized ones and 24.4% among large enterprises. Among large enterprises, the rate rose from 20.4% to 24.4% in a single year. The report explains large companies' heavier use of AI by their capacity to manage complexity, achieve economies of scale, bear the costs and provide the necessary expertise. This picture suggests that RPA gains traction earlier in organizations with high transaction volumes and repetitive processes.
The breakdown by sector follows a similar logic. In 2025, the usage rate of workflow automation was 23.7% in information and communication, 14.9% in electricity, gas and steam, and 12.8% in professional, scientific and technical activities, while it stood at 3.9% in manufacturing and 2.0% in construction.
The effect of scale is clear in Türkiye as well. According to TÜİK data, in 2026 the rate of AI use was 12.8% among enterprises with 10-49 employees and 37.1% among those with 250 or more employees. Because the TÜİK bulletin does not give RPA as a separate figure, it is not possible to compare RPA rates between the EU and Türkiye; the European data should be read here as a directional reference.
In which processes is it used?
Eurostat also asks enterprises that use AI what purposes they use it for. In 2025, 31.1% of EU enterprises using AI applied it to business management and the organization of administrative processes, and 23.1% to accounting, control and financial management. The most common purpose is marketing and sales, at 34.7%; logistics, at 6.1%, is the least common area. Among the example purposes, the report also lists invoice processing.
These ratios apply to all AI technologies, not to RPA alone. Even so, the fact that administrative and financial processes rank so high shows that the first stop for automation is mostly the back office. We covered which finance processes deliver returns faster in our article on finance automation.
A practical path: how to choose your first RPA process
The steps below are a general framework derived from the findings in the sources; each organization needs to adapt it to its own process structure and systems.
- Build a candidate list from the back office. Administrative work, accounting and finance processes stand out in the European data. List repetitive tasks such as invoice entry, reconciliation, report compilation and document classification.
- Measure the volume and the rules. For each candidate, record the monthly number of transactions, the time spent, and the error and rework rates. The high level of use among large organizations indicates that volume is the deciding factor.
- Separate the "rule" part from the "judgment" part. Fully rule-based steps can be solved with classic automation. Steps that require reading free text, recognizing documents or classifying, on the other hand, call for an AI component; the category Eurostat measures is precisely this combination.
- Define the human approval points up front. Put in writing in advance which amounts above which threshold, or which types of exception, will go to human approval. The responsibility for an automation that provides decision support still rests with the process owner.
- Set the success criteria before the pilot. Record the current values of metrics such as time, error rate and exception rate, and compare them against the same metrics at the end of the pilot.
- Plan maintenance for system changes. Bots must be updated when the screen, file format or regulation they depend on changes. The process owner, the technical owner and the change schedule should be clearly defined.
- Build capability within the team. Eurostat's report shows that the most frequent reason given by enterprises that have considered AI but do not use it is a lack of expertise (70.3%). Training the team that will run the pilot is a precondition for scaling.
- Look at the function, not the "agent" label. Automation products are increasingly marketed as agents. When evaluating, focus less on the product name and more on which step it actually runs autonomously; we covered this topic in detail in our article on agent projects.
Conclusion
The European data says two things at once: AI-powered RPA is still limited across the general population of enterprises, but it is spreading quickly among large companies and finds the most traction in administrative and financial processes. The fact that AI use in Türkiye almost doubled in a year shows that the same question is on the agenda for organizations in Türkiye as well.
The starting point should therefore be the choice of process, not the choice of technology. A small pilot run on a high-volume, measurable back-office process with a clearly identified owner provides the data needed to decide on scaling at the lowest risk.
Where logistics experience helps. In the Eurostat breakdown of purposes above, logistics comes last at 6.1%, yet the habits that logistics operations run on (a measured baseline, a few KPIs, written exception rules) are exactly what steps 2, 4 and 5 ask for. One of our consultants brings more than 30 years in operations, logistics and supply chain: at a technology distribution company of roughly one billion dollars in scale, he ran logistics, warehousing, technical service and administrative operations end to end, building and leading teams of 100+ people and delivering lasting gains in operational performance through KPI-driven programs. He has also managed nationwide warehouse and transport operations. An industrial engineer, he now brings that supply-chain discipline to SMEs through AI and automation. Meet our consultants.
#interimmanagement #forwarddeployedengineering