Table of Contents:
Last week, two assessments emerged from the European Commission and the Organisation for Economic Co-operation and Development: the annual report under the Digital Decade, which includes reports for each member state, including Romania, and the Digital Government Outlook. Following these key documents, the temperature gauges of digital transformation worldwide, Romania finds itself at a (new) critical juncture. As digital transformation is like a building marked for seismic risk, in danger of collapse at every political tremor, it remains to be seen how badly damaged its structural integrity will be by what is happening in this period.
With the formation of the new Government, we cannot help but ask what is desired on this subject and how the proposed vision aligns with European directions and those of OECD member countries. I deliberately used this passive construction of the verb “to desire” because, until now, responsibility for action in digitalisation has remained unassigned. At a formal level, we have never lacked a figure on whom this portfolio has been placed: from ministers to deputy prime ministers and (more recently) to proposed delegated ministers and agency directors. What was missing? The political legitimacy and human resources to prioritise, direct, lead, align, harmonise, monitor, and evaluate what happens in digital transformation and, more recently, in artificial intelligence.
Drawing on the Digital Government Outlook and Romania’s Country Report under the Digital Decade, we examine in detail the global trends and where our country stands in this ecosystem.
Governance as the Common Thread and Solution to Institutional Fragmentation
If there is one common thread running simultaneously through the Commission’s report, the OECD report, and all Edge Institute research over the past year, it is governance. For the first time since these assessments have been carried out under the Digital Decade framework, governance is mentioned five times in the Commission’s country report, which explicitly identifies the need for a “comprehensive digitalisation strategy backed by firm political will, sustainable post-RRF financing, a coherent regulatory framework, solid ICT expertise and a well-defined governance structure” (p. 2), going further to stress that it is crucial that “this governance structure be designed for institutional continuity, protecting long-term digital objectives from short-term political cycles.” (p. 27)
The OECD diagnoses the same governance problem at the global level: “The 2025 Digital Governance Index shows that the digital governance of governments is strongly oriented towards planning: strategies are solid, mandates are clearer than ever, and ex-ante evaluation is widespread. What remains weak is the capacity to iterate, evaluate, and adapt once delivery has begun.” (p. 24)
For Romania, which does not even have an updated digitalisation strategy, only a Digital Decade roadmap that has not been revised since 2023, a subject we have written about at length last year, the OECD’s conclusion can only alarm us: if countries with good strategies fail at implementation, a country without a clear vision, direction, and decision-making architecture faces a significantly higher level of risk.
“Governments do not lack individual initiatives, policy tools, or digital components. What they lack is the institutional coherence that makes these elements reinforce one another.” (p. 23), the OECD emphasises. And further: “Data systems that cannot be shared, infrastructure built but not adopted, investments approved but unevaluated, artificial intelligence deployed but unmanaged, and services designed but unconnected. Each of these is a symptom of the same underlying condition.”
Fragmentation is reiterated by the European Commission, which adds that each institution has preferred to invest in its own siloed solution. The direct consequence, as seen from the Commission’s perspective: “Romanian institutions continue to be unable to meet their own national legal obligations regarding the application of the once-only principle” (p. 20), that principle which stipulates that the citizen must not provide the state with data it already holds.
So, while the strategy discussed above is a programmatic document, what we lack, both before and after its adoption, is coordination at the central level. Together with Digital Nation, in October 2025 we proposed a four-level governance model familiar to our regular readers: a Deputy Prime Minister for Digital Transformation with a strong political mandate; a Government Chief Information Officer (GCIO) for professional continuity; an inter-ministerial Digital Council with binding decisions; and an agile implementation agency. The chain of command must be short and clear, under the same institutional umbrella.
The Commission’s 2026 report validates, at least partially, the same model, recommending that the future strategy be coordinated “at a high political level, for example by Romania’s Deputy Prime Minister responsible for digitalisation.” (p. 27)
The OECD gives concrete examples of consolidated governance at the central level, such as the United Kingdom, which in January 2025 merged the Government Digital Service, the Central Digital and Data Office, and the AI incubator into a single Government Digital Centre “with a mandate covering strategy, delivery support, and AI governance across all departments, an approach designed to reduce the distance between central direction and day-to-day delivery.” (p. 26) The global trend is the strengthening of digitalisation governance and its placement close to the centre of government. Countries have understood that they must close ranks, and Romania must not be the exception to this global direction, one no longer recommended only by Edge Institute and the OECD, but also by the European Commission, and validated by the approaches adopted by other states, including some that are digitally more mature.
More specifically, governance is a subject that cannot be avoided even when discussing artificial intelligence or specific projects such as the European digital identity wallet, as we reiterated in “Two Steps Taken, Five Steps to Go”, and as the authors of the aforementioned reports also note: “The development of the European digital identity wallet solution is at an early stage, with the necessary regulatory governance framework still missing and the technical solution yet to be developed” (European Commission, p. 24) and “Governance tools must be actively deployed, continuously maintained, and integrated into daily decision-making in order to generate sustained improvements in AI maturity and digital resilience” (OECD, p. 112).
Artificial Intelligence and Romania’s Ambitions Not Backed by Implementation Measures
Romania speaks of opportunities and regional ambitions in the sphere of artificial intelligence. The discourse is, however, disconnected from reality on the ground: we have an economy that has only very tentatively begun to adopt this technology, and a state that has not clearly defined, at an operational level, its role with regard to AI, nor the concrete measures through which it will reach the level of ambition promoted publicly. This structurally consequential transition must be underpinned, alongside the governance mentioned above, by institutional capacity, skills, guidance on usage, investment, and applied measures.
In practice, although Romania has an AI strategy, implementation has remained an aspiration even at less than a year from the strategy’s expiry — which indicates that we find ourselves precisely in the trap the OECD described: states that have strategies but lose their way at the implementation, monitoring, iteration, and course-correction stage. It is not surprising that the OECD places Romania among the countries that have no designated institution for implementing their AI strategy, given that 83% of OECD member countries and 67% of candidate countries have at least one institution responsible for governing AI use in the public sector, and “these bodies play central roles in developing standards, ensuring ethical oversight, guiding procurement and risk management practices, and translating high-level intentions into operational decisions.” (p. 115)
A nuance worth mentioning: between the OECD’s data collection period (January 2023–December 2024) and today, Romania has adopted a memorandum covering supervisory and notification responsibilities in the context of the AI Act, a limited mandate, insufficient in relation to the complexity of the subject it seeks to address.
Integrating AI into digital public services is synonymous with the state’s proactivity in its relations with citizens; a dimension on which Romania scores 0.06 out of 1 in the OECD’s assessment, a score that approaches the default mark, in a landscape where the OECD member average is 0.67.
One direction France has chosen to pursue, and which could also work for Romania, is to equip, rename, reorganise, and build the capacity of the Authority for the Digitalisation of Romania with competences and resources on the operational side of AI use in public administration.
Tasked to Perform, Left Without the Instrument: Digital Transformation and AI Integration Without the Necessary Human Resources
To be or not to be? To build or to buy?
Two existential questions, the latter of which the state often answers with “to buy” when it comes to digitalisation projects. As the OECD also stresses, the problem arises when outsourcing exceeds a certain level: outsourcing without sufficient internal capacity to govern it is a form of institutional abandonment: “If governments outsource too much, they may lose the internal knowledge and judgment needed to define what they truly want, assess whether they are getting what they ordered, hold providers accountable, and adapt systems when circumstances change.” (p. 95)
We all know the story of the number of specialists at ADR, a two-digit number, with 1 as the first digit. There is a vicious cycle in the Romanian state’s technical capacity: the state cannot attract good technical specialists because it cannot or will not pay competitively; without good internal specialists, projects are poorly conceived and supervised; poorly supervised projects deliver sub-optimal or fragmented solutions; sub-optimal solutions fail to demonstrate the value of digitalisation and reduce political appetite for sustainable investment in human capacity; and reduced investment perpetuates the inability to attract specialists. The fragmented landscape, dependence on external contractors, and the low adoption rate of digital services by citizens are all consequences of the same mechanism.
The way out of this cycle is through a political decision to treat the state’s internal technical capacity as a matter of national security and sovereignty. As the OECD also stresses, internal technical capacity does not mean that the state must hire teams to write code from scratch for all its IT systems — it means the state must have, within its institutions, sufficient qualified specialists to fulfil four essential functions that cannot be delegated to a private contractor without losing control over its own systems:
- Architecture definition and standards. The state needs experienced systems architects to define reference architectures, interoperability standards, security requirements, and design principles that all procured solutions must comply with.
- Informed technology procurement. Terms of reference for complex IT systems must be written by people who understand the technology at sufficient depth to specify requirements precisely, detect exploitable ambiguities, and critically evaluate received technical bids. A poorly written terms of reference is a poor contract, and a poor contract is a guaranteed source of additional costs, litigation, and solutions that do not perform as expected.
- Technical supervision and acceptance. During implementation and at acceptance, the state needs people capable of independently evaluating whether what has been delivered corresponds to what was ordered, in terms of real technical quality, security, and performance.
- System maintenance and evolution. Any complex IT system requires, after implementation, updates, bug fixes, adaptations to new legislative requirements, and continuous improvements. If the state has no people who understand the internal architecture of its own systems, every modification becomes a new procurement exercise or a perpetual dependency on the original contractor.
“Across the spectrum of possibilities, the correct answer is not choosing one approach over the other, but finding a deliberate balance. Governments need sufficient internal expertise to clearly define their needs, procure responsibly, manage suppliers effectively, and maintain oversight throughout the duration of a contract. External suppliers can then complement this internal capacity, being used strategically for specific competences or capabilities, and not as a substitute for the core expertise that governments themselves should hold.
As AI adoption accelerates, the question of what governments should develop internally and what they should procure externally is becoming increasingly pressing. AI systems require ongoing oversight, adaptation, and governance, not just an initial deployment. Governments that lack internal AI expertise risk adopting systems they do not fully understand, cannot evaluate effectively, and cannot adjust when they produce poor or harmful results.” (OECD, p. 97)
The formation of the new Government can be an opportunity to address the structural problems set out above. On this occasion, we reiterate the five steps we proposed in the second call for digital transformation:
- Reconfirmation of the Deputy Prime Minister position for digital transformation, with a clear mandate dedicated to accelerating the digitalisation of public services.
- Appointment of the ADR president as Government Chief Information Officer (CIO), with the rank of Secretary of State (a rank already carried by the ADR president position), at the helm of a reformed ADR placed at the Centre of Government: apolitical, with a clearly defined technical mandate and solid organisational capacity in its reformed structure.
- Operationalisation of the Digital Council through the restructuring of CERB (the Committee for Efficiency, Resilience and Budget) and the expansion of the role and structure of that body’s Consortium, through a monthly schedule of mandatory meetings and the adoption of binding executive decisions, alongside the passage of the legislation necessary to operationalise this measure.
- Launch of a dedicated Government Digital Wallet team. A unit of at least 20 specialists with a single leader, inspired by the Swedish model (which will be further illustrated in a forthcoming Edge Institute study based on exchanges of best practices with the Swedish Agency for Digital Government), as a first pilot of the model. The success of this working model must form the basis for reorganising ADR from an intermediary between European funds and the administration into an orchestrator of public service digitalisation, through the creation of a Digital Delivery Unit built from this dedicated team, serving as an implementation structure for digital initiatives and reforms across central public administration. The precedent exists in Romania: a delivery unit with international expertise operated at the Victoria Palace in 2014 and contributed to the launch of ANAF’s Virtual Private Space (SPV).
- Building AI governance and usage capacity. AI is already being used at scale without any clear picture of actual adoption in government or education, without guidelines, without a framework, and without consistent operational capacity. Five deliverables: a mandatory guide on AI use in public administration adopted by Government Decision; an AI procurement framework integrated into the National Public Procurement Agency’s (ANAP) procedures; at least two pilot programmes in public services with public reporting and efficiency indicators; the inauguration of the national AI regulatory sandbox before August 2026 (as required by the EU AI Act); and a national AI literacy programme for public administration. Compliance with EU obligationsimplementing legislation, ANCOM becoming fully operational, and Digital Omnibus deadlines — remains the minimum mandatory threshold.
Between governing and governance, between priorities and the state’s capacity to deliver on what it sets out to do, and between choices about mechanisms, roles, and responsibilities — the ordinary citizen can only hope that the long-promised digitalisation will eventually reach them too. And proactive public services, if we dare not ask for too much.
10 insights: European Commission, Digital Decade Country Report - Romania (2026)
- Romania has addressed only 40% of the 15 recommendations issued by the Commission in 2024, 50% of the 8 recommendations from 2025, and has established only 11 national targets, of which 36% are aligned with EU 2030 objectives.
- Romania leads the EU in fixed connectivity (VHCN coverage 95.9%, FTTP 95.7%), but 5G stagnates at 46.8% versus the EU average of 94.3%.
- Only 27.7% of the population has basic digital skills, last place in the EU, versus the European average of 55.6%. The gap is most severe in rural areas (20.99%) and among older people (6.17% for the 65–74 age group).
- Digital public services are growing rapidly (score 62.75, a 20.2% increase since 2023), but remain below the EU average of 82.3. Only 26.91% of the population actually uses these services, versus the EU average of 74.71%.
- The digital public services landscape has developed in a fragmented way, each institution built its own siloed solution, and institutions do not comply with the once-only principle, despite it being enshrined in national legislation.
- Only 3.1% of companies use AI (EU average: 13.5%), 15.5% use cloud (EU average: 39%), and 21.9% use data analytics (EU average: 33.3%). Romania has no unicorns in the digital sector.
- Romania trains many ICT specialists (6.9% of higher education graduates), but fails to retain them. The number of employed specialists stagnates at 2.8%, and the elimination of fiscal incentives for the sector is worsening the situation.
- R&D expenditure stands at only 0.52% of GDP in 2023 — among the lowest in the EU. ICT R&D spending constitutes 23.86% of total business R&D expenditure, also among the lowest in the EU.
- Romania has allocated 22% of its recovery and resilience plan (5.8 billion EUR) to digital and has an active portfolio of major projects. Real impact is not expected until after 2026.
- Political coordination at the highest level; for example, by a dedicated Deputy Prime Minister, for the implementation of a digitalisation strategy must be a priority.
10 insights: OECD, Digital Government Outlook (2026)
- Governments have built digital foundations (shared infrastructure, interoperable systems, open data frameworks), but the challenge now is to move from these foundations to real transformational impact for citizens and businesses.
- Only 63% of public institutions in OECD countries actually share data through national interoperability systems, a critical component for proactive services. Privacy standards exist almost universally, but operational data management practices are not keeping pace.
- Digital identity is widely available, but adoption remains low. The challenge is the trust and ease of use needed to prompt citizens to actually adopt it.
- Governments plan and fund digital projects better. Most evaluate projects before launch and allocate dedicated funding, but post-implementation evaluation is quasi-absent: only one in four OECD states systematically checks whether completed projects have delivered the promised results.
- Phased digital funding with incremental learning produces better results than monolithic funding. Governments adopting this approach adapt more easily to new technologies and avoid costly mistakes locked into large contracts.
- Digital skills among civil servants are an ignored crisis: only six OECD states have a dedicated strategy for developing these skills. Without the right people, governments cannot oversee responsible AI use and cannot sustain long-term digital reforms.
- AI is already used in at least one government domain in almost all OECD states, and most have dedicated strategies, oversight bodies, and training programmes. The shift from experimentation to systematic integration has begun.
- Practical support for AI procurement is available in just over half of OECD states. Without this support, governments poorly manage supplier relationships, data rights, and long-term costs.
- Only 28% of states systematically evaluate the impact of AI use in government. Without this measurement, there is no capacity to distinguish promising pilots from solutions ready for large-scale deployment.
- User-centred public service standards are widespread but applied inconsistently. Only 28% of states systematically measure the administrative burden that services impose on users, and feedback loops for continuous improvement remain the exception.
