In legal work, the value of artificial intelligence (AI) lies less in raw speed and more in traceability and risk management. Tasks such as legal research, document comparison, contract review and internal policy audits are areas where a well-designed tool can materially speed up preparatory work and narrow the scope for mistakes.
According to Megyeri Andrea, head of innovation and content development at Wolters Kluwer Hungary Kft., solutions built specifically for legal workflows — such as Libra — differ from chatbot logic: it is not only the quality of the answer that matters, but also which source database the system uses and how it makes its reasoning auditable. This becomes crucial where many documents must be reviewed and compared in a short time and errors can have direct professional or compliance consequences.
AI’s adoption in legal practice
Over the past two years AI has moved beyond theoretical discussion in the legal profession: the arrival of ChatGPT noticeably accelerated interest and adoption. The Wolters Kluwer “A jövő jogásza” research shows that 92% of respondents now use at least one AI tool in their daily work.
That said, fully integrated, workflow-level use is still emerging. Megyeri Andrea estimates that AI currently contributes roughly 5–15% of daily legal work, while in the short term 30–40% of legal tasks could be supported—especially lower-value, repetitive, preparatory activities.
The research also finds that generative AI helps lawyers save about 6–20% of weekly time. This lower-than-possible figure reflects both underutilisation of capabilities and differences in how usage is measured.
Where AI is used today
Common applications remain well-defined sub-tasks: legal research, summarising long texts, translations and drafting initial versions of emails, memos or internal summaries. In many organisations AI functions as an assistant at the periphery of processes rather than being embedded in core workflows.
AI can give a real advantage where document volumes are large, repetition is high and preparatory work is time-consuming. For mass document handling — for example in a large transaction or a housing development with hundreds of contracts — a legal AI can flag deviations, risky clauses and gaps so lawyers can focus limited expert time on substantive intervention.
Professional control and source criticism are the risks
Megyeri emphasises that caution in legal practice is not simply a generational trait: because law is a trust profession, speed alone is insufficient — answers must withstand professional scrutiny. A major risk with generic tools is that users cannot see which sources the system relied on: generic models can produce inaccurate or fabricated citations, conflate current and historical legal states, or simply lack high-quality Hungarian legal training data. These are professional risks, not mere inconveniences.
Another danger is that users treat AI outputs as final without verifying sources. In law it is necessary not only to know the answer but to know where it came from. This is also a training issue: junior lawyers must learn to trace the path from source to conclusion.
Therefore legal education should incorporate AI use as a core competency — not just a technical sidebar — under the headings of professional control, source criticism and verification logic.
What distinguishes legal-specific systems?
A legal-specific, source-verifiable system ties outputs to concrete legal authorities and identifies citations at paragraph, subsection and date levels. Libra, for instance, is built on a database that has been developed and maintained for over 30 years; the system shows what it worked from and allows users to drill down into sources for further research.
Libra integrates research, document analysis and drafting on a single interface, producing auditable results and a less fragmented workflow.
Agentic approach versus chatbot
A chat model is essentially reactive: it answers the question asked and produces a single output. The agentic approach, by contrast, accepts a task or goal: it creates a multi-step plan and works with specialised sub-agents (one searches, another drafts, a third verifies or retries). In practice this means it orchestrates a workflow rather than merely providing an answer.
This difference matters because legal matters rarely stop at “what is the rule?” One must research, organise information, prepare an internal briefing, draft client-facing communication or an initial document, all while keeping sources traceable. Agentic AI arranges these interrelated subtasks. Libra follows this logic by treating legal research, analysis and drafting as connected parts of a single workflow.
Use case: mass document handling and compliance
Mass document handling is one of the most tangible use cases. Manually reading large sets of documents is slow and monotonous. An AI system can ingest all documents, apply review criteria and highlight deviations, risky points and omissions. That does not replace lawyers, but it quickly identifies where substantive professional intervention is required.
A more everyday example: a junior researcher might previously have spent up to 10 hours on a research question from which only 4–5 chargeable hours could realistically be billed. If that work can be completed in 1–2 well-structured hours today, the operational difference is significant. On the corporate side, a HR professional can prepare standard employment contracts or job descriptions faster and with better control and less external input, which has tangible value.
Compliance is another strong application area. When regulation changes, AI can compare current laws and commentaries with internal policies and flag where updates are needed. This accelerates policy revisions, focuses internal controls and shortens post-incident compliance reviews.
Impact on billing and the legal service model
If research and document preparation take a fraction of the time, freed capacity must be allocated: more clients, more complex matters, strategic work or business development. In practice the work mix changes rather than the overall workload: less mechanical, repeatable work and more substantive legal thinking and client advising. On the corporate side, faster handling of standardisable documentation yields direct cost savings.
Market rollout: Libra acquisition and launch
Wolters Kluwer announced the acquisition of Libra on November 14, 2025. Within months, by the end of March 2026, Libra was rolled out in nine European markets; the Hungarian launch took place on March 26, 2026. That pace indicates real market demand for legal AI workspaces of this kind in at least some regions and customer segments.
The Libra proposition combined an established technology with an organisation-maintained legal content base, creating a solution that preserves professional control while accelerating workflows.
Conclusion
The ultimate value of legal AI is not speed alone but how well it maintains professional control and makes outputs traceable to sources. The most effective tools do not take responsibility away from lawyers; they help practitioners work faster while keeping results auditable and professionally defensible.
The publication of this article was supported by Wolters Kluwer Hungary Kft.
Photo credit: Mudra László/Portfolio
