AI strategy
AI for lawyers in Barcelona and Paris: less admin, confidentiality intact
Obinova helps law firms in Barcelona and Paris use AI to take admin hours off lawyers and support staff: appointments, engagement letters, court notices, billing. Legal judgement stays with the lawyer, and the data rules are agreed before anything is built.
Where AI for lawyers saves time first
Many partners hear about AI for lawyers and picture a tool that drafts pleadings. That is the part a firm should be slowest to hand over. The hours come back sooner from the work around each matter: confirming appointments, preparing the engagement letter, logging the deadline on each court notice, pulling the monthly bill together, replying to the client who wants to know how things stand.
Obinova works with law firms in Barcelona and Paris, the two cities where we are based. We go through those tasks with the people who do them, automate the ones that always follow the same steps with n8n, and train lawyers and support staff on assistants such as ChatGPT, Claude and Microsoft Copilot. The approach fits a two-lawyer practice as well as a firm with several practice areas.
A draft, a summary or a date suggested by a tool always goes through a lawyer, who answers for it to the client and to the bar.
There is a practical reason for that. In the Harvard Business School and BCG study of 758 consultants, people using AI produced work of over 40% higher quality and finished 25% faster. Outside the tool's strengths, though, results got worse. In a law firm, review is what covers that gap.
What we do for your firm
You can take on one of them or combine them, depending on what the firm needs.
An audit of how the firm runs
We follow a matter with partners, associates and support staff, from the first call to the final bill. We rank the tasks where AI helps and write down the data rules. Each mission opens with a kick-off session and continues with reviews several times a year.
Automation with n8n
n8n is a tool that connects applications and chains automatic steps together. We use it to link email, calendars, spreadsheets and, where it allows, your practice management software. Workflows are designed with your team and documented so your staff can change them without relying on us.
Training for lawyers and support staff
ChatGPT, Claude, Gemini and Microsoft Copilot applied to real tasks: summarising a file, comparing two versions of a contract, drafting a client letter in English, Spanish or French. It covers which data may go into which tool. In person at your office or on our video platform.
Six law firm tasks and what automating each one depends on
We rank them on three criteria: how often they repeat, what data they handle and whether the firm already has a written template. The first ones can start quickly. The last ones need firmer rules, or stay with an assistant used by a lawyer.
Appointments and first consultations
A workflow confirms the appointment, sends a reminder the day before and a short form about the type of matter. A shared calendar is enough. Little data is involved and none of it very sensitive, which is why this usually comes first.
Conflict checks
Using the opposing party's name from the form, the workflow searches your client and matter records for matches and alerts the lawyer in charge. It only works if those records are tidy. Whether there is a conflict is always a lawyer's call.
Engagement letters and money on account
From your templates, the workflow fills in the engagement letter with the client and matter details and prepares the request for money on account. You need an up-to-date template for each type of matter. The lawyer checks the fees before anything is sent.
Court notices and deadlines
When the firm downloads a notice from the court platform (LexNET in Spain, the RPVA in France), AI summarises it and suggests the deadline for the diary. A lawyer checks it before it counts, every day, with no exceptions. This is where a mistake costs most, so it comes after the tasks above.
Time recording and bills
The workflow gathers recorded time, diary meetings and emails for each matter and prepares a draft bill or time report. It relies on the team recording time reasonably consistently. The partner in charge signs off the final amount.
Status updates for clients
An assistant drafts an email from the file notes explaining where the matter stands and what happens next. It needs clear notes and a tool approved for client data. The lawyer reviews it and sends it under their own name.
Confidentiality and GDPR: the rules we agree before building
In Spain and in France, lawyers are bound by professional secrecy over everything they learn through their work. Putting matter data into an AI tool means entrusting it to an outside provider. That is why these rules come before any workflow.
Only tools with a data processing agreement
The provider signs an agreement under Article 28 of the GDPR, does not train its models on your data and says where the data is stored and for how long. Free personal accounts are ruled out for any client data.
The minimum needed, pseudonymised
Summarising a judgment or comparing clauses rarely needs the client's name or ID number. They are replaced with references before the text goes into an assistant.
Extra care with health and criminal data
A personal injury claim contains health data, a special category under the GDPR, and criminal offence data has its own rules. For those matters the firm decides case by case which tool may handle them, and sometimes the answer is none.
Individual accounts and a log of each workflow
Everyone signs in with their own account, and n8n workflows record which data passed through each step. If a client asks what was done with their information, the firm can answer.
Your bar and your data protection officer for specific questions
The Barcelona bar (ICAB) or the Paris bar, and your data protection officer if you have one, are the reference for doubtful cases. We help you prepare the questions with the details of each tool.
How we work with a firm, from first call to follow-up
- 01
Free 30-minute video call
Send us a message through the contact form and we will email you to agree a time. On the call you tell us which tasks weigh most and which software the firm uses.
- 02
Written proposal
We send the scope, the calendar and the quote. You can start with the audit, a single workflow or the training.
- 03
Kick-off with partners and support staff
At your office in Barcelona or Paris, or by video call. We follow a typical matter from start to finish and agree the data rules.
- 04
A first workflow on a small scope
Appointments and first consultations, for example. Before starting we note a simple indicator, such as the support hours spent on that task, to compare against later.
- 05
Training and follow-ups through the year
We train the team on the firm's own cases, anonymised beforehand, and review results several times a year to decide the next step.
Two cities, three languages, and what the cost depends on
Firms in Barcelona often work in Spanish, Catalan and English in the same week, and Paris firms with international clients switch between French and English. An assistant used well prepares a first draft in the client's language, and the lawyer checks the legal terms, which machine translation does not always get right. We work in Spanish, English and French, which helps firms with clients or correspondents on both sides of the border.
Kick-offs and in-person training take place at your office in either city. Workflows are mostly built remotely.
We don't publish prices. The quote depends on how many tasks are reviewed, how many tools need connecting in n8n, how many people are trained and how often you want follow-ups. After the first meeting we send a written proposal with the scope and the price.
Frequently asked questions
How much does AI cost for a law firm?
It depends on scope: the number of tasks reviewed, the tools to connect, how many people are trained and whether training is in person or on video. Training a three-person practice is a very different job from automating admin across several practice areas. We send a written quote after a free 30-minute video call.
How long does it take to get the first workflows running?
The calendar is set out in the proposal and depends on the task chosen and the firm's software. We start with one tightly scoped task, such as appointments or engagement letters, and check it against your indicator before going further. Training can run in parallel. Follow-ups through the year show the effect over several months of data.
Is AI compatible with legal professional secrecy?
Yes, with written rules. Only tools with a data processing agreement and no training on your data are used, the minimum goes in and it is pseudonymised, and health or criminal data is handled case by case. A lawyer reviews every output. For specific questions, check your bar's guidance and ask your data protection officer.
Is it worth it for a small practice?
Often more than for a large one, because in a two- or three-lawyer practice the lawyers do the admin themselves. Appointment reminders, engagement letters and bills follow fixed steps that an n8n workflow handles well. Start with one, measure how long it takes before and after, and decide on that basis.
Can a chatbot answer my clients' questions about their case?
It can answer practical questions: opening hours, which documents to bring to the first meeting, how fees are paid. It must not assess a case or give legal advice. Any question on the substance goes to a lawyer, and the assistant tells the client so. The firm approves every answer in writing before it goes live.
Which AI tools can lawyers use safely?
General assistants such as ChatGPT, Claude, Gemini or Copilot help summarise a judgment, compare two versions of a contract or draft a first client letter, provided they run on a business account with a data processing agreement and the data is pseudonymised. Do not rely on them to cite case law without checking, because they can invent references. A lawyer reviews every output.