Your prompts and expertise, turned into firm-wide processes
Your lawyers already use AI, but each in their own way. We turn the team’s best practices into repeatable procedures that work directly with matter documents and can be integrated by every team member. We build primarily on Claude – a leading AI model - and a lawyer always approves the output.
Implementations led by Tomasz Żernicki and Wojciech Ozimek — business process auditors for Wolters Kluwer Poland, founders of technology companies.
Experience behind Legatura
Since 2017, our team has worked with Wolters Kluwer as executive consultants and later as auditors supporting its AI transformation.
“The ability […] to translate complex topics, including AI, into practical and actionable business recommendations — proved particularly valuable.”
An AI licence does not automate a process
Giving the team a new tool rarely changes daily work by itself. An AI assistant must be embedded in a specific part of a process, automate agreed steps and leave responsibility for the outcome with a person.
According to Thomson Reuters' 2025 research, 45% of law-firm respondents were already using generative AI or planned to make it central to their workflow within a year. At the same time, 48% still had no formal GenAI policy.
A senior lawyer spends an average of 30–40% of their time on administration and repetitive work that does not generate revenue. In a 15-lawyer firm, this translates into EUR 12,000–20,000 per month in unrealised revenue or avoidable cost.
If your team already has prompts or instructions that work in chat, we take them further: we turn them into procedures that work on client documents, produce repeatable outputs and can be run by anyone on the team.
Source: Thomson Reuters, 2025 Generative AI in Professional Services
A window of opportunity for firms that move before their competitors
Claude Code, Codex, ChatGPT and Google Antigravity can work with legal documents at a quality sufficient for real support, not just experimentation. Law firms in Western markets have been introducing them into operational work since 2024.
Firms that implement AI in key processes in 2026 can build a 12–18 month operational lead over competitors. AI adoption is no longer a question of whether, but when and how.
Most Polish law firms are still exploring. The first firms to move from experimentation to production implementation will gain a cost and competitive advantage that will be difficult to close later.
Start with work that repeats every week and genuinely takes time
We do not start with a feature catalogue. We select one workflow with an owner, available data, a measurable result and a sufficiently large loss of time or cost.
The AI assistant prepares a summary of the enquiry, conflict-check data and a matter-opening checklist. A team member verifies the package.
The AI assistant prepares a current view of the matter from agreed documents and correspondence. A lawyer approves the brief.
The AI assistant transfers client data into an approved document template, which is then reviewed by a lawyer.
What the work looks like before and after implementation
A specific example: a matter brief and status update — one of the most common tasks in a law firm.
These figures are indicative. We determine actual results during the Process Map using baseline measurements from your firm.
For every process, we choose the simplest effective approach: implementation, integration or custom development. If an existing tool solves the problem better, we implement it instead of building a custom assistant.
A procedure works on a folder of documents, not a single file in chat. Where dozens of contracts, a year of correspondence or hundreds of compliance documents must be reviewed, standard tools hit context limits. Our procedures iterate through documents, and repeatable elements move into dedicated tools.
Where to start depends on where you are today
Every firm is at a different stage of AI adoption. The table below helps identify the most effective next step.
| Stage | Where the firm is today | Recommended first step |
|---|---|---|
| 1 | The team does not use AI or uses it occasionally, without agreed rules or designated owners | AI Assistant Workshop |
| 2 | The team uses AI and has its own instructions, but outputs are not repeatable, transferable or usable across multiple documents | Process Map |
| 3 | The firm knows which process to improve, has documentation and an owner, and is ready to implement | AI Implementation Sprint |
Not sure which stage your firm is at? We will establish it during a 20-minute call.
From the first assistant to an AI-supported process
Hands-on work, not a presentation: a partner and 1–2 selected team members build a working assistant for a low-risk task using the firm's materials. You leave with a working assistant and the ability to build more.
Together with a firm partner, we identify the processes that consume the most time, examine one with the person who performs it and use the Pareto method to select the workflow to implement. You leave with a map of one process, the recommended workflow and a fixed price for the Sprint.
We implement AI in one agreed workflow, test the solution with users and measure the time recovered.
After implementation, we monitor adoption, KPIs and subsequent technology decisions.
This workshop format has already been tested with a demanding audience
We use the same format with product and project managers at technology companies. The law-firm version works with legal documents and processes.
"To me it's always best to learn new technologies by having someone experienced present them from a practical point of view. And this is exactly what this workshop was about. Pure practical approach. No BS. Hands-on stuff."
"This was not another 'future of AI' session. It was six hours of practical work […] I strongly recommend this workshop to anyone who wants to stop treating AI as a toy and start using it as a real operating system for day-to-day work."
Technology companies are already building their own assistants. Law firms that act now will be among the first in their market.
One outcome, the right engine
An agentic tool from the creators of one of the leading AI models. It works directly with the firm's files and folders, executes multi-step procedures and captures the team's way of working in plain language. Where the client's environment requires it, we use alternative platforms listed in the FAQ.
For the partner who wants an outcome, not an experiment
Confidence in the decision
Before approving an AI implementation, a partner should have clear answers to several questions.
The Process Map is an independent diagnosis. It selects the best process based on evidence, not on what is easiest to sell. If the diagnosis shows that the firm is not ready for a Sprint, we will say so.
We process firm data only with Business or Enterprise-grade tools that provide data isolation and GDPR compliance. Configurations, rules and client data belong to the firm. The methodology and solution architecture remain our know-how, just as with professional advisory work.
The Map gives the partner concrete figures: how much time the process takes today, how much can realistically be recovered and what the Sprint will cost. No guessing. The decision is based on evidence, not enthusiasm.
“My job is not to make technology look impressive. It is to find where it produces a measurable result.”
We combine management, process and technology perspectives
Our team is led by Dr Tomasz Żernicki — founder of the international technology company Zylia, contributor to MPEG Audio standards and experienced in building and scaling technology businesses in the US, Europe and Asia.
Before we talk
Claude Code, Codex and Antigravity offer similar agentic capabilities but differ in environment, permissions, integrations and operation. We select one platform that fits the task, data and users.
We use only tools selected for their documented compliance with applicable regulations and data-protection requirements. Every implementation takes into account data risks, security and usage rules. However, compliance also depends on the firm's specific context, configuration and use of the system. We do not provide a blanket legal-compliance guarantee for the entire firm. Where necessary, the client should involve their Data Protection Officer or legal counsel.
No. We design processes so that AI output is reviewed and approved by an accountable person.
No. The assistant performs preparatory work as a defined procedure: it gathers data, organises documents and prepares a draft. The expert decides whether to use, revise or reject it. AI supports the team; it does not replace professional judgement.
A Sprint covers one workflow: a clear start, end, owner, data, output, one team, no more than two core systems and one primary time KPI. We split larger processes into separate workflows.
We do not quote a number before diagnosis. During the Map, we review documents, conduct interviews, measure the baseline and define a realistic success criterion. After the Sprint, we compare completion time with the starting point.
The Map is a standalone service. The firm keeps the recommendations, roadmap, KPIs and implementation specification.
The Map remains a standalone service. You keep the process register, recommendations, roadmap and quotation. There is no obligation to continue. The partner decides on evidence, not promises.
We do not promise a specific number of hours before measurement. We promise three things: we measure time before, measure it after, and if the result does not confirm savings, diagnose the cause and plan a correction at no additional cost. The partner is never left alone with a solution that does not work.
During a 20-minute call, we will determine whether there is a suitable first workflow and whether we are a good fit.