Over the past three years, Ukraine has been quietly executing a massive institutional transformation: a comprehensive Public Investment Management (PIM) reform. Undertaking such an overhaul would be a significant and complex logistical challenge for any nation under normal circumstances. Ukraine, however, is managing this nationwide reform while enduring hundreds of Russian drone and missile strikes targeting its cities every week. International financial and institutional estimates currently place the total cost of rebuilding the country at well over US$ 588 billion, spread across thousands of distinct projects involving local communities, regional authorities, national ministries, and dozens of international donor partners working simultaneously.
At the epicenter of this ambitious reform is DREAM, Ukraine’s open digital ecosystem designed specifically for public investment management. Built from the ground up on core open government and open contracting principles, DREAM serves as a central technological bridge. It connects state-owned and external registries, supports continuous project monitoring and sophisticated data analysis, and creates a single, transparent digital pipeline for public investment projects across the nation.
Today, DREAM acts as the digital backbone of the entire PIM reform. It currently supports more than 12,900 public investment projects spanning 18 distinct sectors. The platform is actively utilized by more than 1,300 municipalities, every regional state administration, all government ministries, and major state-owned enterprises.
The ecosystem covers the entire investment lifecycle of public works, ranging from initial project conception and feasibility appraisals to formal prioritization, funding coordination, and implementation monitoring. As this foundational reform transitions into its full-scale implementation phase, Ukrainian authorities and their international partners are introducing artificial intelligence as a practical enabler. The goal is to strengthen institutional capacity, make intricate PIM regulations usable for every municipality in the country, and ultimately improve project quality, planning efficiency, and delivery speed.
The Scale of the Challenge
This unprecedented scale brings forth an immediate, practical capacity challenge for public officials. Administrative staff across the country must navigate a labyrinth of complex laws, rigid methodologies, public procurement rules, construction norms, financial requirements, and strict donor conditions.
Local governments, in particular, frequently operate under severe resource constraints, featuring limited staff, fragmented expertise, and very little time to pore over hundreds of pages of legal texts, methodologies, and operational manuals. For instance, completing a full cycle of an investment pre-feasibility study for a relatively small municipal project typically demands about four weeks of full-time, dedicated work from a single local government team. Consequently, preparing just ten projects simultaneously becomes an overwhelming administrative hurdle for a small local government office.
During peak planning and submission cycles, the DREAM Project Office handles more than 850 support requests per day, even with extensive training programs, active helpdesks, and expert networks already in place. During 2025 alone, the project administration organized 96 training events that engaged nearly 30,000 participants, alongside dedicated helpdesks and technical support networks. Despite these extensive efforts, traditional institutional support models struggle to scale at the rapid pace demanded by the reconstruction process.
This ongoing capacity shortage is made significantly more acute by the relentless pressures of the ongoing war. More than 230 communities, representing roughly 15% of the country, hold front-line status. These areas face daily artillery shelling, dense drone attacks, severe energy infrastructure breakdowns, and massive waves of internally displaced persons (IDPs) seeking refuge. All of these front-line communities must still be fully included in the national PIM process and retain equitable access to critical reconstruction financing.
When administrative capacity is stretched this thin, even sound and well-designed regulatory rules become exceptionally difficult to apply in daily practice.
Our Approach to AI
Artificial intelligence cannot replace public servants, expert human judgment, or political accountability—at least not in the foreseeable future. However, it can dramatically improve how vital knowledge, including policies, methodologies, regulatory requirements, and sectoral expertise, is accessed, interpreted, and applied. This is particularly crucial within a fast-evolving and highly complex reform environment. Bridging this specific operational gap is the primary objective of the new technological initiative.
Within the DREAM ecosystem, artificial intelligence is not viewed as a flashy external automation tool. Instead, it functions as an integrated support layer embedded directly into daily governance workflows. Its fundamental purpose is straightforward to explain but challenging to execute: to help municipal and national users prepare and appraise better projects, understand regulatory requirements faster, and make more consistent, evidence-based decisions.
Three years of intense reform implementation have provided the project with a solid foundation that most artificial intelligence initiatives typically lack. The initiative relies on a verified, comprehensive knowledge base, a clear mapping of where users routinely encounter bottlenecks, and the hard-won operational data necessary to build a genuinely useful assistant. Rather than serving as a superficial marketing feature, the tool acts as a functional mechanism designed to support and empower the people tasked with rebuilding the nation.
With strong political backing from the Ministry for Development and the Ministry of Economy, the DREAM Project Office joined forces with the Open Contracting Partnership, which brings recognized expertise in providing AI tools for public procurement. They partnered with uStudio as the primary IT and AI innovation partner, alongside GCIEP and PwC, with support from the UK Foreign, Commonwealth & Development Office (FCDO), to develop an AI-powered assistant seamlessly integrated into the DREAM ecosystem.
This collaborative approach is grounded in a tested working prototype, a curated and verified knowledge base, and a hybrid architecture strictly aligned with Ukraine’s national data sovereignty requirements and the ethical AI principles outlined in the official White Paper on Artificial Intelligence Regulation in Ukraine.
The DREAM AI Assistant: Three Stages of Implementation
The rollout of the DREAM AI Assistant is strategically aligned with the broader PIM Reform Roadmap spanning from 2024 through 2028, and it advances through three distinct operational stages.
During the initial phase in 2026, the AI assistant provides real-time guidance on PIM rules and the project preparation process directly within the established DREAM interface. The primary objective here is to bridge the wide gap between complex legal language and practical municipal application. The assistant achieves this by translating dense methodological requirements into accessible, clear explanations tailored specifically to each user’s individual role, administrative sector, and geographic location. Serving as a trusted, round-the-clock entry point into the comprehensive PIM knowledge base, the assistant delivers practical guidance for government bodies at every administrative level.
Initial testing of the DREAM AI Assistant took place in April 2026. Focus group participants representing various local communities tested the assistant throughout the project preparation process, awarding it an average performance score of 4.33 out of 5 points. While developers continue to tailor and upgrade the tool based on this feedback, it has already demonstrated its ability to reduce administrative workloads and accelerate procedures by guiding DREAM users smoothly through project preparation.
Looking ahead to 2027, the DREAM AI assistant will expand its functional scope into quality assurance. It will assist appraisal teams in identifying missing indicators or internal inconsistencies in feasibility studies, while proactively suggesting ways to better align proposed projects with national sectoral priorities. This phase will also focus on improving the economic, environmental, and financial cases of submitted proposals. By comparing new project submissions against historical, validated cases, the system aims to enhance the overall consistency and transparency of thousands of impending investment decisions.
In the 2028 to 2029 timeframe, the strategic focus will shift toward advanced investment modeling. Utilizing geographic, demographic, fiscal, and other socioeconomic data, the system will help public officials explore various development scenarios and model alternative investment options. This advanced functionality will enable rigorous, scenario-based assessments to determine precisely where public resources can achieve the greatest possible societal and economic impact.
Building a Safe and Responsible AI Architecture
Because public investment decisions involve exceptionally high stakes and sensitive administrative data, the DREAM AI assistant has been constructed upon a hybrid, fully auditable architecture engineered to guarantee responsible governance.
The core of the technological system utilizes Retrieval-Augmented Generation (RAG). This ensures that every generated response is grounded exclusively in verified national laws, operational manuals, and official guidelines. To safeguard sensitive public information, the platform employs local language models hosted securely on protected infrastructure, thereby upholding national data sovereignty. Every answer provided by the system is fully traceable to an approved source, timestamped, and version-controlled. Furthermore, the underlying architecture integrates a mandatory human-in-the-loop validation process to ensure that all generated content remains accurate and reflects the most recent regulatory updates.
To minimize the inherent risks of artificial intelligence hallucinations or the accidental use of outdated guidance, every system-generated response links directly to an approved source and carries a specific timestamp and version identifier. Potential operational risks, such as factual inaccuracies or data privacy concerns, are proactively mitigated through structured prompt design and rigorous access control protocols.
A dedicated multidisciplinary oversight team continuously monitors the system to ensure it adheres strictly to established ethical standards and remains firmly focused on its core public purpose.
Ukraine’s Lesson for the World
Ukraine’s ongoing experience demonstrates that effective implementation of artificial intelligence in government does not begin with complex algorithms. Instead, it starts with clear rules, robust institutions, and the disciplined design of technology that directly serves the public interest. The artificial intelligence layer currently being integrated into the DREAM ecosystem is only as effective as the reform infrastructure beneath it—an institutional foundation that required three years of intense labor and immense dedication to construct.
If executed successfully, this pioneering initiative has the potential to become not only a vital catalyst for Ukraine’s national recovery and long-term development, but also a globally relevant case study illustrating how artificial intelligence can responsibly strengthen public investment management systems within highly complex environments. Proving what is possible while rebuilding a country under fire, supported by cutting-edge digital governance, establishes a powerful precedent for administrative resilience worldwide.
