Table of Contents:
The workplace
In the third wave of the Edge Institute survey conducted by AtlasIntel we added a question about respondents’ occupational status, and this allows us, for the first time, to analyze the answers about artificial intelligence through the lens of Romanians’ workplace. The sample is divided into 3 groups: 45.1% say they do not work, 42.5% work in the private sector, and 12.4% in the public sector.
Thus, the place where a person works largely determines whether and how artificial intelligence is used, and the results show us that the state must both serve and regulate an economy that is rapidly adopting AI, as well as understand where guidance, education, and rules are needed for the use of AI within the state.
Sample structure by occupational status
"Which of the following statements applies to you?" .% of respondents
The engine: the private sector
Employees in the private sector are the group most connected to artificial intelligence: 87.9% use it for at least one purpose (personal, educational, professional), while those who do not use it at all stand at 12.1%. Compared to the public sector, the differences are multiple: an overall use nearly 30 percentage points higher (87.9% versus 60.6%), as well as double the use at the workplace, where 29.5% of private-sector employees already use AI tools, compared to 16% in the public sector.
AI use, by work sector
"Do you use artificial intelligence tools?" .%
Click on an occupational status to see how the use of artificial intelligence varies depending on the workplace.
The private economy therefore appears to have already integrated AI, independent of any public policy: the transformation seems to have begun from the bottom up, through the initiative of people and companies.
Thus, the productivity gain appears where the use of artificial intelligence is integrated into processes, and the private sector is for now the only one in which use at work reaches a critical mass close to a third of employees. If the advance remains concentrated here, the gap between the firms that adopt AI and the rest of the economy, but also between the economy and the administration, will deepen.
The brake: the public sector
The most relevant contrast for public policy emerges from the public sector, where 39.4% of employees say they do not use artificial intelligence at all (not even in their personal life), more than three times more than in the private sector, with use in any form standing at 60.6% of respondents. At work, only 16% declare that they use such AI tools, almost half compared to the private sector. In other words, the workforce from which the state’s capacity to use AI in institutions should be built appears itself to be the least familiar with these tools, taking into account, of course, the possible reluctance to acknowledge such use in the absence of institutional norms.
AI knowledge level, by work sector
Of those who use AI %
However, analyzing the public-sector employees who already use AI, their skills are comparable to those of the other respondents: 58.3% place themselves at an intermediate level and 10.7% at an advanced one, slightly above the private sector. The barrier therefore does not seem to be related to skills or reluctance to learn, but to the absence of organized exposure and of a framework that encourages the use of these tools in administrative work.
But precisely because adoption starts from a lower level, state interventions can produce more prominent results, and citizens’ openness toward the use of artificial intelligence in administrative tasks (analyzed in the previous newsletter) gives the state (still) a mandate to turn its attention to AI in the public sector, which public sector therefore needs decisions that transform individual/sporadic use into a guided, controlled, and educated institutional practice.
Frequency of use
Once the threshold of adoption at work is passed, the differences between employees in different sectors fade. Of those who use AI, the share of those who use it at least weekly is similar across all three groups: 76.6% in the private sector, 80.2% in the public sector, and 76.1% among those who do not work. It is important to emphasize that this frequency is a combined one, covering personal, professional, and educational use together, and not only use at the workplace. Likewise, the level of knowledge declared by users is comparable, with advanced-level shares of 9.3% in the private sector, 10.7% in the public sector, and 12.6% among those who do not work.
How often they use AI, by sector
Total frequency of use (any context, not just at work), among users %
Those who do not work
The group of those who do not work, the largest in the sample, reveals a distinct profile: personal use is high, at 77.7%, close to that in the private sector, but the level of competence is the lowest (62.5% of users place themselves at the basic level). Within the category there are, of course, different realities, from students who use AI at school (12%, with 12.6% advanced users) to pensioners and people outside the labor market, with a more superficial exposure.
This is precisely the picture that reveals a widespread use of artificial intelligence, but a use that is overshadowed by fragile competencies acknowledged by the users themselves (about which we wrote more in Romania in the age of artificial intelligenceand in the report on Automation in Romania), and it is yet another alarm signal and yet another proof of the need for digital literacy and AI-use programs, addressed to the general public.
Broken down, students would need the integration of AI into education, with rules for correct use, while people looking for a job could benefit from reskilling programs that transform superficial use into professional competencies. The window for launching such programs must be now, in the present, before technological advance and the qualification of only minority segments of the population deepen the gap between citizens even further.
The presence of young people in this group nevertheless deserves a special note, because the students who use artificial intelligence in the educational environment are also the ones who most often declare a (more) advanced level, a sign that early exposure could produce increased competencies. The same reality shows, however, that the education system must already manage the use of these tools, with all that they imply for assessment and for academic integrity.
Fear, inversely proportional to exposure
The most apparent tension emerges from the fears for one’s own job, because although at the national level only 21.9% of respondents declared themselves worried about their own job, the sectoral breakdown reveals that this average includes diametrically opposed positions.
Thus, public-sector employees are the least worried that AI could affect their job in the next five years, with only 7.7% worried or very worried and 40.2% not worried at all, while private-sector employees are more concerned (18.2%), and those who do not work express the greatest unease (29.5%). The order therefore seems, on the one hand, to be inverse to exposure: the group most protected by job stability is also the least worried, while the most vulnerable group, those without a job, is the most anxious. On the other hand, the ranking seems directly proportional to vulnerability.
The fear that AI could affect their job, by sector
"How worried are you over the next 5 years?" %
Despite the public discourse, young people are alarmed about the accelerated integration of AI into the labor market. 1 in 2 respondents in the category aged between 18 and 29 expresses worry or even severe worry on this topic.
The explanations are multiple, and they do not necessarily relate to the person’s socio-economic or occupational status. Romania is already the EU member state that reaches the highest rates of young NEETs (young people not included in a form of education or work activity) annually, a rate standing at 19.2% of the country’s young population in 2025. The prospect that automation takes over jobs and replaces human intelligence thus creates a fear of growth in the NEET category, but also an important dilemma regarding the role of young people in a society in which AI becomes a dominant factor.
The fear that AI could affect their job, by age category
"How worried are you over the next 5 years?" %
What is to be done, starting from the state
The sectoral analysis thus reveals that, at least as far as the state is concerned, the challenge regarding artificial intelligence is first internal and only then external, because the administration cannot offer citizens AI-supported services as long as its own employees use it the least and do not benefit from guidance or direction.
Viewed through the work sector, the map of artificial intelligence in Romania thus reveals a private economy that has already set out on the paths of the new AI era, a large group outside the labor market that uses the technology without mastering it, and a public sector left behind both in adoption and in awareness of the change.
Sources and references
- AtlasIntel for Edge Institute, „The digitalization of public services in Romania III”, survey conducted between 17-21 June 2026, sample of 2.002 respondents, margin of error of two percentage points, 95% confidence level. Segmentation by occupational status (private sector, public sector, people who do not work).
- Edge Institute, Apel pentru transformarea digitală a României (2026)
- Edge Institute, România în era AI: supraexpuși, nepregătiți, absenți
- OECD, Governing with Artificial Intelligence (2024)
- GOV.UK, AI experiments see „Humphrey” help townhalls cut costs and improve services (2025)
- Comisia Europeană, #Bürokratt national virtual assistant platform (Estonia)
- Amnesty International, Denmark: AI-powered welfare system fuels mass surveillance (noiembrie 2024)
- Eurostat, Statistics on young people neither in employment nor in education or training
The current wave includes 2.002 respondents, data collected between 17–21 June 2026. The comparative analysis integrates the previous waves. All three waves use the same Atlas methodology — random digital recruitment — with a margin of error of ±2 p.p. at a confidence level of 95%.
ⓘ Where the wording of the questions or the list of options changed between waves, the value comparison is indicative.