Contents

Ihor Bilyk, Roman Maselko, Claude

Part I. What has already been done

1. Access to data: a difficult road that took three years

Article 31 of the Law of Ukraine “On the High Council of Justice” contains a strikingly clear guarantee of the Council’s institutional capacity: members of the HCJ have the right of direct access to state information resources — without a court order, without any permit procedure.

The legislator did this not out of excessive generosity, but because the HCJ decides on the appointment of judges, the dismissal of judges, disciplinary liability and temporary suspension. Each of those decisions affects both a specific person and the independence of the judiciary as a whole.

At first we, too, read Article 31 narrowly: as an individual guarantee for a Council member. It quickly became clear that what is at stake is the informational foundation of every function of the High Council of Justice — disciplinary, personnel-related, investigative, analytical. Without it, the Council works not with facts, but with whatever was sent to it.

Processing information has to begin with putting the mechanisms for obtaining data in order. The first attempts to obtain data from the State Geocadastre, the register of real property rights, the Unified State Register of legal entities, the register of encumbrances on movable property, enforcement proceedings, the Unified State Demographic Register, the State Register of Individuals, the register of vehicles, criminal-record data and a number of others were made in September 2022. The most common argument for refusal ran as follows: there is a Regulation on the relevant register, approved by ministerial order; it contains a list of users; and the HCJ is not on that list.

We are not inclined to see ill will here — rather inertia. Every data controller guards its own register. The incentives are arranged so that caution is always more advantageous than cooperation. There is only one way to change this: to make electronic interaction the normal state of affairs and a paper request the exception.

Digitalising the process of obtaining data revealed a rather interesting effect, a kind of “digital progression effect”: digitalising the internal processes of one institution indirectly triggers the digital transformation of processes in other bodies it interacts with.

Today, obtaining data is no longer handled by an individual member but centrally by the secretariat of the High Council of Justice, under an established procedure, with registers prioritised and a clear roadmap in place. The High Council of Justice finally has a comprehensive picture of which data its members, disciplinary inspectors and secretariat staff need in their work.

The Unified State Register of Court Decisions is a story of its own. Formally, the HCJ had access to it. In practice, full access — with complete texts of decisions and the ability to run systematic searches — was held only by a few officials. Yet the assistants to HCJ members, that is, the people who actually worked through the materials of disciplinary cases, had no such access. Imagine reviewing a complaint against a judge without being able to read the full text of a court decision. That is not a hypothesis; that was the working regime.

We will not dramatise or ask how the Council managed before. We will put it more moderately: a significant part of the work was carried out under informational constraints that objectively reduced both the speed and the completeness of reviews. For the first time, full access was granted on a large scale not only to HCJ members but also to secretariat staff and disciplinary inspectors.

What did this achieve? Registers stopped being an exotic resource. They became a daily instrument: Council members obtained the data required for their powers, dependence on external correspondence decreased, and reviews became more complete. The High Council of Justice has access to 9 registers. This is not a revolution. It is the precondition without which nothing that follows would have made sense.

2. The analytics unit

Access to information does not by itself create capacity. The next step was therefore to establish an analytics unit — a division that centrally gathers information from registers, works with resources requiring secured automated workstations, conducts OSINT and GEOINT research, verifies and cross-checks information from different sources, and prepares analytical materials. Thanks to the effective organisational work of the head of the unit, Olena Atai, staff moved into the field of digital research considerably faster than expected.

This is an important conclusion, and it runs counter to a widespread belief. To build a workable analytical division inside a state body, you do not have to spend years hunting for narrowly specialised technical experts with market-level salary expectations. What you need is: a correctly defined direction, access to basic digital tools, methodological support, and a clear understanding of what practical result is expected of you. At present, the core work of collecting, initially verifying, systematising and interpreting data is done by a team of five.

Previously this work was done by HCJ members, their assistants and other secretariat staff — each for themselves, in parallel, duplicating one another. Centralisation freed up the working time of Council members and of more than sixty assistants. It freed it up for what they are there for: the legal assessment of facts and the preparation of opinions and decisions.

The price of optimisation

Now for the part that articles like this usually leave out. Centralisation created a new problem. Five people carry a volume of work that used to be spread across dozens. They do it efficiently — and that is exactly where the danger lies. The efficiency of five easily turns into an argument against further development: if they are coping, why change anything?

In fact the opposite is true. Concentrating a large volume of manual work in one small unit is not an achievement but a precise diagnosis. It shows what has to be automated next, and it shows it down to the individual operation.

For example, if 60–70% of the time is spent transferring data, then that is the stage to automate. By our estimates, it is realistic to automate 80–90% of data-processing operations. That figure is made up of concrete steps, and we describe them in the sections below.

The effect of such automation is not only speed. It also means a lower probability of a typo, an arithmetic error or a missed material circumstance. One item is hard to measure but impossible not to name: less overtime for people who have been working for a fifth year under massive shelling and chronic sleep deprivation.

3. D-3: the tool that was there all along

The story of digitalisation at the HCJ is closely bound up with software that is old and utterly familiar to the judiciary — the “D-3” computer program, and specifically its templating functionality, which has been there since “time immemorial”. The tool that was needed had in fact existed long before digitalisation became mainstream.

The developers of the “D-3” suite created a universal templating mechanism at the time. It allows you to: create templates for any type of document; pull data from the system into a document automatically; return information entered in a document back to the database; use conditional constructions; generate a document taking into account the details of a specific case; sign the result with a qualified electronic signature; and automatically register the document and add it to the case file.

This is, in essence, a ready-made generator of procedural documents that had been working for years. And for years only a handful of people used it. Iryna Mamontova, Larysa Shvetsova, Anatolii Miroshnychenko and Olena Blazhivska deserve mention as the first to apply it in practice. Yet the practice never became systemic and disappears when the terms of office of the HCJ members who used it come to an end.

This is a clear demonstration that the problem of digitalising state bodies rarely lies in the absence of technology. Far more often it lies in the fact that nobody treats the existing technology as a task in its own right: there is no process owner, no templates, no training, no requirement to use it.

Fifty templates and one ruling

With the intention of changing this, more than 50 templates for various types of documents were prepared in just 2 months and made available to any user: rulings, requests, letters, cover notes and other procedural and organisational documents.

The most telling example is the ruling returning a disciplinary complaint where formal grounds exist. The document is standard, the reasoning is standard, the volume is small. Preparing it used to take one to two hours; with a template it takes two to five minutes.

The difference is not that people started typing faster. The working scheme itself changed: the standard reasoning is already in the template, the case details are pulled in automatically, the complainant’s data is not re-entered by hand, the document is signed with a qualified electronic signature and is an original from the moment of signing, and it is registered and added to the electronic case file. The whole chain disappears: print it out; catch an HCJ member between sessions; sign it on paper; register it; scan it; upload it back — plus a series of other unnecessary steps.

The Second Disciplinary Chamber is piloting it

The Second Disciplinary Chamber of the High Council of Justice was the first — and so far the only — body to adopt this approach in its work. Its decisions are now drawn up and signed electronically.

Tellingly, disciplinary inspectors took to the new practice most easily. The reason is simple: they started working after the electronic tools appeared and had no habit of orienting themselves towards paper. Nobody taught them “how it used to be”.

This suggests a strategy: retraining is harder than training. Every new process is worth launching as paperless from the outset, rather than digitising it later.

We expect this practice to spread to the other disciplinary chambers, to HCJ sessions, to other categories of proceedings and to the internal administrative processes of the secretariat. Only then will the separate electronic procedures finally add up to an end-to-end digital process.

4. A new position: CDTO

Digital projects have one unpleasant property: they do not survive without a permanent “project owner”. As long as digitalisation is an additional burden on a few motivated people, it moves at the speed of their free time and stops when they take a holiday or move to another job. This is why a separate position has appeared in public administration — deputy head for digital development, digital transformation and digitalisation, or CDTO for short.

Why is it needed if there is an IT division? Because these are different things. An IT division keeps the equipment and systems running. A CDTO is responsible for something else: architecture, priorities, the link between technical change and real business processes, coordination between the HCJ, the High Qualification Commission of Judges, the State Judicial Administration, the state enterprise “Information Judicial Systems” and donors — and for making sure that ten projects run by different international partners do not turn into ten incompatible systems.

At the time of writing, the appointment of a CDTO at the HCJ is being decided. This is probably the key personnel precondition for everything discussed in Part II.

5. The UJITS: a story a decade long

The development of the Unified Judicial Information and Telecommunication System (which, after being renamed, becomes the Unified Judicial Information and Communication System, UJICS) is the longest and hardest strand of judicial digitalisation. The reasons accumulated over years:

• the HCJ does not have enough technical specialists of its own;
• the State Judicial Administration’s capacity to design complex information systems independently is limited;
• dependence on external technical assistance;
• the termination of US government programmes and of USAID activity;
• the prospects of funding from the state budget remained uncertain;
• responsibility between the HCJ, the State Judicial Administration and the technical administrator is divided in such a way that each is responsible for a part and nobody is responsible for the result;
• agreeing technical and functional requirements took months.

How things got moving

The unblocking point was the Memorandum of Intent signed on 23 October 2025 within the project “Strengthening the European dimension of the Ukrainian civil service in the fields of lawmaking, justice and public finance”, in the part concerning the digitalisation of the judiciary and constitutional justice. The Memorandum was concluded between the civil society organisation “Laboratory of Legislative Initiatives” and the Supreme Court, the High Council of Justice, the High Qualification Commission of Judges of Ukraine and the Constitutional Court of Ukraine, under the auspices of the Swedish International Development Cooperation Agency (Sida).

Under the project, technical documentation is being developed that is intended to become the basis for new software products for the judicial system, including the technical requirements for UJICS subsystems. The Memorandum expressly provides that it creates no legally binding rights or obligations, yet its signature by the Chair of the HCJ creates entirely real institutional and reputational expectations. All the more so because the project’s results are tied to implementation of the Rule of Law Roadmap and the Ukraine Facility Plan for 2024–2027 — that is, to Ukraine’s commitments on its path to the EU.

The quality of the technical documentation determines everything: whether future solutions will be compatible with existing systems, whether they can be integrated, operated safely and maintained, and whether they can be taken over by the State Judicial Administration at all. Four sets of technical requirements have already been prepared and approved by the State Judicial Administration.

The interim conclusion on the UJITS strand is that the preparatory stage has been restored. But the real measure of success will not be a concept, a roadmap or a set of technical specifications — it will be a working subsystem used by judges and parties to proceedings.

6. Artificial intelligence and real projects

Over the past two years, more has been said about artificial intelligence in Ukraine’s legal community than has been done. So the limits the HCJ will never cross should be stated up front. This is not about replacing a judge, an HCJ member or a disciplinary inspector with an algorithm. It is about far more prosaic things: finding relevant case law, classifying documents, identifying similar legal relationships, summarising large volumes of data, analytics, detecting patterns and anomalies, drafting supporting materials, and reducing repetitive manual work.

In the short time since 16 July 2021, when the day of artificial intelligence was officially introduced, the HCJ has already done substantial organisational work:

• Analysis of case law to draft court decisions — a joint project with the Ministry of Digital Transformation, the State Judicial Administration and the National Police.
• A judge’s assistant for searching case law — a joint project with the Ministry of Justice, the State Judicial Administration, the state enterprise “Information Judicial Systems”, individual educational institutions and outside experts. It has two strands: searching case law on similar legal relationships and — far more interestingly — identifying systemic problems in the application of law and assessing the effectiveness of legislation. The second strand means being able to see which provisions consistently generate disputes, and to see it from a body of decisions rather than from expert intuition.
• Analysis of the HCJ’s disciplinary practice — a joint project of the HCJ and the Better Regulation Delivery Office (BRDO). This is the project closest to us. It makes it possible to search for relevant decisions, harmonise approaches, see recurring types of violations, analyse the duration and outcomes of proceedings and prepare statistics. And, most importantly, to make disciplinary practice more predictable.

The internal organisational processes of the High Council of Justice concerning staff use of artificial intelligence for everyday routine tasks deserve separate mention. An anonymous survey we initiated showed that 88% of HCJ secretariat staff have experience of using artificial intelligence, and 72% already apply that knowledge in their own work.

Why this is moving so slowly

There are many reasons, and all of them are genuine: the difficulty of agreeing changes within a large institutional system; the sensitivity of judicial data; the absence of common standards; limited funding; a shortage of computing and human resources; concerns about errors, bias and legal liability; and the need to guarantee that the final decision is always taken by an authorised human being.

Experience in other digital areas shows a regularity: as soon as a clear methodology, adequate infrastructure and user trust appear, isolated experiments turn into stable products very quickly. The problem is almost never the model — it is that the first three conditions are missing.

One practical solution deserves a separate description, because it removes the biggest objection: data sensitivity. It also illustrates how the seemingly most difficult question — privacy — can be resolved without any risk at all. The most concentrated form of private data is the “analytical report” on candidates. Yet by technical, purely local means these data can be de-identified before being passed to a model. Under this approach, personal data are replaced with artificial values (a hash or invented data). The de-identified document is then passed to the model for analysis. Having received the result, the program performs the reverse substitution: the artificial values are replaced with the real ones. The model never sees genuine personal data, and the user gets a complete result. De-identification, incidentally, does not require artificial intelligence: regular expressions and local named-entity recognition handle it on their own.

This is exactly the kind of solution usually missing from debates about AI in justice: not “may we or may we not”, but “how exactly do we make it so that we may”.

Part II. What cannot be left undone

What follows is not a wish list but a concrete vision of what the High Council of Justice must do within defined deadlines. The digitalisation of the HCJ must be set out not only at the level of national or international measures to be implemented, but must also be firmly fixed in the body’s own documents.

• The HCJ Digitalisation Strategy for 2026–2030 — a philosophy of digitalisation from which no departure is possible.
• The internal automation plan for HCJ structural units — the most concrete of them. It is written not in the language of goals but in the language of operations: which unit, which process, how long it takes, exactly what is being automated, and which ready-made tool already exists.

The most underrated category of measures is those that require no changes to legislation, no large procurements and no other body’s consent. What they require is a few weeks of work.

The analytics unit: a report that fills itself in

Four registers have already been automated: the State Register of Individual Taxpayers, the “Arkan” system (data on border crossings), the Unified State Register of Vehicles and the Unified Register of Powers of Attorney. Next in line are the Register of Real Property Rights, the State Land Cadastre, NACP declarations, an individual’s court cases, access logs of the Unified State Register of Court Decisions and others.

Automating the processing of data from these registers will make it possible to compile an analytical report 90% without staff involvement. On top of that comes functionality that simply makes the work convenient: supplementing information from open sources; an interactive interface for viewing without duplicating data; and a single materials viewer.

The session preparation unit: time spent inefficiently on session materials

Staff of this unit spend a substantial share of their time creating individual folders, structuring them and renaming files. Automating this process requires neither legislative change nor millions: export the materials by a list of case numbers together with a “file — case” mapping table, then sort and rename them automatically according to that table, shortening the names, including with the help of artificial intelligence.

Next comes something that should change the work of Council members: a streaming materials viewer. Instead of downloading a folder with dozens of files — a single continuous stream that scrolls like a PDF, with the next document loading as the next page. A prototype has already been implemented as part of the work on the Electronic Court. In coordination with the state enterprise “Information Judicial Systems”, the viewer can be integrated with D-3 directly, so that materials do not have to be downloaded each time.

An interactive session calendar, in which staff themselves add items for consideration and the system, drawing on voting records from previous periods, predicts the time needed for each item. The consequence is mundane, but for anyone who has attended a session it is very tangible: instead of “everyone at 10:00” — specific times and intervals.

The back office: de-identification without manual work

Publishing decisions today runs up against manual de-identification. Detecting personal data can be automated by local means: regular expressions plus named-entity recognition. The effect should be faster publication of decisions.

The disciplinary inspectors’ service: manageable case-law summaries

A tool for managing summaries of HCJ practice generated by artificial intelligence (developed jointly with BRDO) will make it possible to change the approach to compiling annual summaries and turn them into a living instrument. All summaries can be produced with the help of artificial intelligence under the control of a responsible person, and users will gain not only speed but also the ability to track, interactively, how the approaches of the HCJ and its bodies to particular questions change over time.

The strategic frame: digitalisation as a measurable quantity

Quick wins produce an effect within months. But without a framework they remain a set of useful utilities. That is why the draft HCJ Digitalisation Strategy for 2026–2030 is built around an idea that is rare in Ukraine’s public sector: digitalisation is measured.

The yardstick chosen is the CMMI process maturity model. The logic is that digitalisation is not the number of systems purchased but the manageability of processes:

Level	What it means for the HCJ	Deadline
2 — managed	Key processes are documented, have owners, and results are tracked	Q4 2026
3 — defined	All processes are described in a single register, standardised and carried out electronically	Q4 2027
4 — quantitatively managed	Every process has quantitative speed and quality targets; indicators are measured automatically; deviations are detected statistically	—
5 — optimising	Processes are continuously improved on the basis of data; the effect of every change is measured	Q4 2030

The most interesting part is how the Strategy extends the fifth level. The function of analysing processes and proposing improvements is to be assigned to artificial intelligence, leaving humans the exclusive right to decide on changes. The cycle closes:

• data — event logs, calendars, analytical systems and the UJICS give a continuous picture of how processes actually run (process mining);
• analysis — the model builds the actual rather than the declared process diagrams, and finds bottlenecks, redundant approvals, uneven workloads and scope for shortening deadlines;
• proposal — with a projected effect: by how much the deadline will shrink, which resource will be freed up, how much fewer errors there will be;
• decision — taken by the process owner; nothing changes without human approval;
• measurement — the actual effect of a change is measured automatically and fed back into the cycle.

The difference between this and a conventional “business process audit” is frequency. An audit is done once every few years by consultants who describe processes as staff described them to them. Process mining looks at the logs and sees how things actually happen. Continuously.

The Strategy’s key caveat: the elements of this cycle must be launched not in 2029 but from the very start, beginning with small areas, so that mistakes are found while they are still cheap.

The rest of the Strategy rests on eight strands of internal digitalisation: paperless business processes; artificial intelligence; data and analytics; digital competences and personnel policy; information security; digital resilience; infrastructure and resource efficiency; and digital communications.

Here are a few target indicators that give a sense of the scale: 100% of disciplinary proceedings in electronic form, 100% automated de-identification of decisions, an 80% reduction in the time needed to compile the agenda, 80% of regular analytical products generated automatically, and an RPO of no more than 15 minutes and an RTO of no more than 4 hours for critical services, confirmed by testing.

In a country at peace, the last indicator would look like a technical detail. In a country where infrastructure is a target for missiles, it means that the body which appoints and dismisses judges must be back at work within 24 hours even after a direct hit.

One strand deserves separate mention, because it is usually absent from the strategies of state bodies: building an internal capacity to develop software using the body’s own staff and AI tools. The experience of automating the work of the HCJ’s analytics unit shows that this is not a utopia but entirely feasible. And that it is an order of magnitude faster and cheaper than a procurement procedure.

It would be dishonest to end a list of plans without a list of what could stop them.

• Money. Some measures require almost nothing; others require serious funding. The Strategy’s answer is prioritisation, frugal solutions (upgrading existing equipment instead of replacing it, open-source software), and partnership and grant programmes.
• People. IT and data specialists on a public-sector salary are in short supply and will remain so. The answer is in-house training, in-house development using AI, and drawing on the expertise of donor projects.
• Coordination. The greatest risk, as the history of donor projects confirms. No internal HCJ plan can compensate for a decision not taken in another body.
• Resistance to change. Not sabotage, but habit. Someone who has worked with paper for twenty years does not become a supporter of paperless processes because of an order. The answer is training, digitalisation ambassadors, taking digital competences into account in personnel procedures — and time.

An unfinished beginning

We called this text “an unfinished beginning”, and we want to explain why that, rather than “a finished” one.

Formally, the preparatory stage really has been completed. The HCJ demonstrated the need for access to data and obtained part of the critically important access rights. It created an analytics unit that works. It began systematically templating documents. It moved individual proceedings into electronic form. It put in place the personnel preconditions for transformation. It resumed work on the UJITS. It joined the first applied AI projects. It wrote a strategy with measurable indicators instead of a list of intentions.

Each of these results is real. But taken together, they are not yet a system.

An analyst has access to nine registers but still spends two thirds of the time moving data by hand. Templates cut the preparation of a document from two hours to three minutes, but not everyone uses them. Case materials exist in electronic form, yet 80% of a unit’s time goes on renaming files.

That is the mark of incompleteness. Not an absence of results, but the fact that they are unconnected. We have a set of well-made parts and, as yet, no mechanism.

That is precisely the next task: to assemble the parts into an end-to-end model in which data arrive automatically, documents are generated from structured information rather than from copied text, deadlines are tracked by the system rather than by an individual’s memory, and reviews follow a single methodology rather than the diligence of one particular person. And in which staff do what they are paid for: assess, rather than transcribe.

Then the beginning can be considered finished. Not yet. And that, paradoxically, is the best news in this text: everything that has not been done is not a dream but a list of tasks with deadlines, responsible owners and working prototypes.

Making it finished is not a question of technology. It is a question of decisions.

Disclaimer

Artificial intelligence technologies were used extensively in preparing this article, to work through large volumes of primary documents and to turn individual ideas of the authors into connected text. This was not done arbitrarily but under a specific methodology developed for the purpose — the same methodology for conducting legal research in the judicial system that is set out in the conceptual documents described above.

All assessments, conclusions and formulations belong to the authors. Artificial intelligence generated none of the ideas in this article. Artificial intelligence will replace hands, but it must not replace the head.

We consider it appropriate to say so plainly. First, because we demand transparency about the use of such technologies from others. Second, because this text is itself an illustration of its own thesis.