Your firm has been accumulating expertise for years. Your team takes too long to find it.
RAGnexus is the AI assistant that learns from your firm's work — past contracts, briefs, legal opinions, and client files — and answers your lawyers in seconds, always citing the source document. Without a single file ever leaving the firm.
In October 2024, our opposition brief focused on three arguments: absence of periculum in mora given the defendant's proven assets, disproportionality of the requested measure pursuant to Supreme Court Ruling (STS) 412/2023, and...
Your firm's collective expertise already exists. The problem is retrieving it when needed.
Every mid-sized firm accumulates years of briefs, legal opinions, template contracts, and case files that reflect the firm's way of working. But this knowledge lives scattered across the document management system, emails of the case manager, drafts in local folders, and the heads of senior partners.
Knowledge locked by person, not by firm
When a senior associate leaves or a new partner joins, client-specific or subject-matter expertise goes with them. What was 'in the head of the lawyer handling the case' never actually made it into the firm's system.
Searches that eat up billable hours
An associate takes 30 to 60 minutes to find an internal precedent, draft template, or past brief. In a 40-lawyer firm, that amounts to hundreds of hours per month that are not billed to clients and can never be recovered.
Associates using generalist AI on their own
The reality is that many juniors are already using ChatGPT to write drafts on their own. Without internal policies or audits, they often copy and paste text containing sensitive client data. The ethical and regulatory risk is on the table whether you want it or not.
Growing competitive pressure
Major firms are deploying proprietary AI. Increasingly sophisticated corporate clients negotiate lower rates because they know their advisors are gaining productivity. Mid-sized firms that don't adapt will see their margins shrink.
There is a way to capitalize on your firm's collective expertise — without exposing a single file to third parties.
Your firm with its own proprietary AI. Without sacrificing professional secrecy or data control.
Three steps. Four to eight weeks. Zero exposure of files to external vendors.
01 — We connect your firm's sources of expertise
We integrate RAGnexus with your document management system (iManage, NetDocuments, Kleos, Lexon, Jurisoft, or your custom folder structure), your corporate email, and the external legal databases you already license (Aranzadi, Tirant, vLex, Lefebvre). The assistant works on your firm's materials — not the open web.
02 — We train and host on European infrastructure
Your firm's document base is processed on dedicated European servers assigned to your firm — or in your own on-premise infrastructure if you prefer. The AI model operates on that material, fully respecting the user, department, and client permissions you have already defined.
03 — Your team asks questions just like they would to a colleague
Directly from the browser, Outlook, or the document management system itself. In natural language. Every answer cites the case file, document, page, and exact excerpt — so any statement can be verified with a single click.
The firm's collective expertise, turned into profitability.
An illustrative case: 40-lawyer firm, corporate and litigation practice areas, offices in Madrid and Barcelona.
The starting point
Associates invested between 3 and 5 hours weekly in internal document search that was not billed to clients (part of the firm's 'invisible time'). Partners wrote reports and legal opinions from scratch because they couldn't locate equivalent past work.
What changes after deployment
The team locates any internal brief, clause, or opinion in less than two minutes. New associates produce work aligned with the firm's accumulated expertise from their first week. The firm no longer depends on the specific case manager to reconstruct backgrounds.
| Metric | Before | After |
|---|---|---|
| Average internal search time | 30-50 min | < 2 min |
| Non-billable hours per lawyer/week | 3-5 hrs | < 1 hr |
| Billable hours freed/year (40 lawyers, conservative estimate) | — | ≈ 4,000 hrs |
| Capacity equivalent without hiring | — | ≈ 2-3 lawyers |
| Estimated annual investment in RAGnexus | — | €40,000-€80,000 |
Illustrative case based on public benchmarks of the Spanish legal sector. The actual return depends on the size of the firm, the current recovery rate, and the digitization level of the document management system. We do a specific calculation for your case in the first meeting, with no obligation.
Why not use ChatGPT, Copilot, or Harvey?
Because none are built for Spanish professional secrecy, mid-sized firms, or to work natively with your own accumulated expertise.
| Concept | RAGnexus | ChatGPT Ent | Copilot | Harvey |
|---|---|---|---|---|
| Works on the firm's internal case files | ||||
| Data processing in European infrastructure | ||||
| Deployment option without calling external APIs | ||||
| Cites case, document, and page in every answer | ||||
| Designed for professional secrecy regulated by the General Statute of Spanish Advocacy | ||||
| Support and onboarding in Spanish, team in Spain | ||||
| Cost adapted to firms of 20-80 lawyers | ❌ ($$$$) | |||
| Dedicated deployment in 4-8 weeks | months |
Works on the firm's internal case files
Data processing in European infrastructure
Deployment option without calling external APIs
Cites case, document, and page in every answer
Designed for professional secrecy regulated by the General Statute of Spanish Advocacy
Support and onboarding in Spanish, team in Spain
Cost adapted to firms of 20-80 lawyers
Dedicated deployment in 4-8 weeks
We do not compete with OpenAI, Google, or Anthropic.
We combine the best of their models — or European open-source alternatives — with an architecture that keeps data control, professional secrecy, and compliance entirely in the hands of the firm. You decide how far your data sovereignty goes: from a dedicated European cloud deployment to a 100% on-premise solution on your own servers, with zero external API calls.
Two deployment methods. You choose the level of sovereignty.
Every firm has a different risk profile. We adapt the architecture to your standards.
RAGnexus Cloud
Fast deployment in a dedicated European cloud. Top-tier commercial models (OpenAI, Anthropic) hosted in European regions under enterprise agreements that strictly prohibit the use of your data for training.
- Deployment in 4 weeks
- GPT-4 and Claude models in Azure EU / AWS Bedrock EU
- Contractual guarantee of no-training
- GDPR compliant
- Full audit log of every query
Ideal for corporate, litigation, and general practice firms needing rapid deployment and top-tier model quality.
RAGnexus Sovereign
Total sovereignty. European open-source models (Mistral, Llama) running on dedicated infrastructure exclusively for your firm — or on your own physical servers. No queries are sent to external APIs. None. Highly recommended for criminal law firms and firms handling clients in highly regulated sectors such as banking, energy, or healthcare.
- Deployment in 8-12 weeks
- Mistral / Llama models on dedicated European infrastructure
- Zero dependence on US providers
- Zero exposure to the CLOUD Act
- On-premise option (firm's own physical servers)
- Full audit log + absolute control
Ideal for firms with clients in regulated sectors (banking, defense, energy, healthcare) and for firms that view the highest standard of confidentiality as an essential part of their value proposition.
*Both options comply with GDPR, LOPDGDD, and the duty of professional secrecy. The difference lies in the level of sovereignty and control over the infrastructure.
Designed to comply, not just to impress.
RAGnexus aligns from its very architecture with the specific legal obligations of a law firm in Spain.
General Statute of Spanish Advocacy
Royal Decree 135/2021. Architecture fully compatible with the duty of professional secrecy. No client files ever leave your controlled environment.
GDPR and LOPDGDD
Processing in European infrastructure. End-to-end encryption. Right to be forgotten and data portability built-in. DPA signed with every firm.
EU AI Act
Designed with the EU AI Act regulatory framework in mind, adhering to the specific obligations that apply to systems used in legal contexts.
ENS — National Security Scheme
Architecture alignable with the Spanish ENS for law firms that provide services to the public sector or work with government administrations.
SOC 2 (in progress)
Security and audit controls aligned with the international standard. Type II certification planned for 2026.
CGAE — General Council of Spanish Advocacy
Designed in line with CGAE recommendations on the responsible use of AI in law and professional ethics.
What your team can ask RAGnexus.
Real query examples. Every response comes with the exact reference to the case and source document.
"What earn-out clauses have we negotiated in technology company acquisitions over the last three years? Show me the ranges and the ones that were eventually executed."
"Do we have any appeal briefs against a ruling dismissing preliminary injunctions in unfair competition matters? I need a starting point."
"Summarize our historical stance in collective redundancy (ERE) negotiations for organizational reasons in the retail sector. What severance numbers have closed our collective disputes in the last five years?"
"What have we opined on transfer pricing in holding-subsidiary structures under the scope of the new Global Minimum Tax regulation (Pillar 2)?"
"I want to see all the rent review clauses indexed to the CPI that we have agreed upon in office leases signed after the LAU reform."
"Which partners or associates at the firm have worked on antitrust matters with the CNMC in the pharma sector? Show me the briefs and the team members in charge."
Built in Spain. Designed for Europe.
RAGnexus was born from an uncomfortable observation: Spanish law firms were starting to adopt AI by relying 100% on US-based providers — Microsoft, Google, OpenAI — with everything that implies in terms of jurisdiction, data sovereignty, and professional secrecy.
We built an alternative: an enterprise AI layer that integrates with the best models in the world but leaves control, data, and liability entirely in the hands of the firm. Today we work with mid-sized firms in Spain where confidentiality is not just a preference: it is an ethical duty with real disciplinary consequences.
"We saw too many law firms choosing between adopting AI or protecting professional secrecy. That trade-off shouldn't exist. With the right architecture, it doesn't."
Javier Pérez
CEO & Founder of RAGnexus
Frequently asked questions from partners & CIOs.
It depends on the deployment mode you choose. With RAGnexus Cloud, queries pass through commercial models (OpenAI, Anthropic) hosted in European regions (Azure EU, AWS Bedrock EU) under enterprise agreements that explicitly prohibit the use of your data for training. With RAGnexus Sovereign, we use open-source models (Mistral, Llama) running on dedicated European infrastructure or your own servers — no query is ever sent to external APIs. None.
The architecture is designed so that no client documents ever leave the firm's controlled environment. In Sovereign mode, not even text snippets are sent to external providers. We sign a Data Processing Agreement (DPA) and confidentiality agreement with every firm, aligned with Article 21 of the General Statute of Spanish Advocacy.
Pricing depends on the number of lawyers, deployment mode (Cloud or Sovereign), and document volume. For a 40-lawyer firm, typical ranges are between €40,000 and €80,000 per year, which includes deployment, support, and maintenance. We calculate the exact quote during our first call — with no obligation.
No. RAGnexus is deployed as a managed service. We handle the integration with your SharePoint, Outlook, or document management system, along with indexing, maintenance, and updates. Your team only needs to use the assistant.
RAGnexus Cloud: 4 weeks from signing. RAGnexus Sovereign: 8-12 weeks, depending on the complexity of the infrastructure (dedicated European cloud or the firm's physical servers).
The knowledge remains in the firm's system. RAGnexus maintains granular permissions per user, practice area, and document type. When a lawyer leaves, their access is revoked, but the documents they indexed (which belong to the firm, not them) remain available to the team.
We offer a 6-week pilot with a selected subset of documents and a pilot group of lawyers. If you choose not to continue after the pilot, there is no penalty. We can discuss the terms on our first call.
Currently, we are focused on firms of 20 to 80 lawyers, where document-related pain points are most acute and ROI is clearest. If your firm is smaller, feel free to contact us anyway — we can recommend alternatives or include you in a future tier for boutique firms.
We strictly respect the permission structure you already have defined in your document management system. If a labor associate is restricted from viewing corporate files, they will not be able to query them via RAGnexus either. If there are ethical walls ('Chinese walls') between teams due to conflicts of interest, RAGnexus respects them natively.
This is exactly the question we get from every managing partner. The answer is: the permission architecture operates at the case file and user level. Each user can only query material they already have access to in the document management system. RAGnexus never cross-references or leaks information between conflicting clients.
We closely monitor publications from the CGAE, the Council of Bars and Law Societies of Europe (CCBE), and regional bar associations. If there is a regulatory shift affecting the product, we update the configuration and notify you before compliance becomes mandatory. This is part of our standard service.
Shall we discuss your firm for 20 minutes?
No canned presentations. No sales pressure. In 20 minutes, we will tell you honestly whether RAGnexus fits your firm, your current document management system, and your confidentiality requirements — or if it doesn't.