Industry · Law & Legal

AI for law firms. Faster case work, with confidentiality intact.

We build AI applications for law firms, notaries, tax lawyers and in-house teams. Document review, case law search and case file summarisation, each with EU data residency, source references and an audit trail that can meet the NOvA (Netherlands Bar Association) guidelines.

PracticeDocument review
PracticeCase law search
PracticeCase file summary
PracticeClient query triage

Dutch law firms in numbers.

~18.500
Lawyers registered with the NOvA (Netherlands Bar)
~4.300
Firms in the Netherlands
~75%
Say AI adoption is a strategic priority
High-risk
Legal AI in Annex III of the EU AI Act

Source: NOvA annual report 2024, Wolters Kluwer Legal Industry Survey 2025, EU AI Act Annex III.

Standard packages don't cover what your practice demands.

Most law firms work with a combination of Kluwer Navigator or Lexis for case law, a DMS for case files and a case management package such as Clio or their own middleware. These suites do their core job well, but as soon as you want AI features, such as document review across hundreds of contracts, semantic search across your own precedents or a chatbot for client queries, you run into standard templates that haven't been set up for your firm.

M&A practices need due diligence tooling that flags critical clauses across hundreds of contracts, family law calls for case file summarisation with privacy guarantees, tax lawyers want legislation monitoring per client portfolio, and in-house teams need contract review flows that integrate with their ERP and CLM. None of these flows fits an out-of-the-box package.

At the same time, the market is under pressure. Commodity suites such as Harvey AI, Spellbook, Robin AI and Eve are billed as revolutionary, but their output is standardised: the same model, the same prompts and the same benchmark playbook at every firm that buys them. What sets your firm apart, such as your specific clause style, your own precedent bank and the insight you've built up in a niche segment, doesn't come back in such a suite. What remains is a generic layer that at best speeds up junior work without making the firm strategically stronger.

We build AI applications that fit onto your existing stack: private LLM deployments for sensitive case work and RAG systems that unlock your own knowledge base, always with source citations so hallucination is not a risk. We don't replace Kluwer or Lexis; we build the custom layer on top that sets your specialism apart. For some firms, that means focusing on a single use case, such as a document configurator that generates contracts from a question tree. For others, it is a broader AI platform that supports multiple practice groups.

AI applications by legal practice.

Every discipline has its own document flows, its own risks and its own compliance requirements. For each area of law, we show what we typically build and which models we use.

M&A and corporate

The due diligence phase is where AI delivers visible value fastest. Hundreds of contracts (supplier, lease, IP, employment agreements) have to be reviewed for change-of-control clauses, notice periods and restrictive provisions. A well-trained LLM flow flags these automatically and groups them by risk category.

Our due diligence modules combine Claude with a dedicated RAG layer on your playbook and previous transactions. Your legal team stays in control: the AI does the heavy lifting and delivers a structured shortlist per data room folder. For more detail, see our work on due diligence platforms.

  • Contract review with playbook overlayYour firm's own playbook decides what counts as "red" or "green", not a generic benchmark.
  • Data room classificationAutomatically detect document types and map them to the right due diligence checklist.
  • Clause extraction with source referencesEvery flag links back to the exact paragraph and page, so there are no unsupported claims.
  • Private LLM deploymentClaude or Llama with EU residency, with no training data leaking to public models.

Built to the strictest standards of legal practice.

Confidentiality, the GDPR and the EU AI Act set strict requirements for any AI application in a law firm. We work within that framework from the first sprint.

Duty of confidentiality (Wbo art. 11a)

Case data stays within the firm

We deploy LLMs within your own cloud environment or in a private EU deployment. Case data never passes through a public model provider, and logging is set up so that privileged information is not stored in plain text.

GDPR Art. 35 — DPIA

Privacy by design for client files

Client files almost always contain special category personal data. For every project, we deliver a DPIA document covering data flows, retention periods and data subject rights, ready for your data protection officer.

EU AI Act — Annex III

Legal AI is high-risk

Legal AI applications fall under the high-risk category of the AI Act. We document risk classification, human-in-the-loop design and bias monitoring from the design stage, not as paperwork after the fact.

NOvA guidelines

Rules of conduct for the legal profession

Since 2024, the Netherlands Bar Association (NOvA) has had explicit guidance on the use of AI. Our flows comply with, among other things, the rules on automatic storage of case data, hallucination checks and transparency towards the client.

eIDAS Regulation

Advanced electronic signature

For client agreements, engagement letters and e-signing of contracts, we build eIDAS-compliant signature flows with an audit log and qualified timestamps.

Wwft and KYC

Client due diligence

For firms subject to the Wwft (the Dutch anti-money laundering act) such as notaries, real estate and M&A, we integrate KYC screening, UBO checks via the KvK (Dutch Chamber of Commerce) and sanctions list monitoring into the onboarding process. You get an audit trail per client, with no more loose Excel files.

Seamlessly connected to the legal tech ecosystem.

We integrate with the legal software stack your firm already uses. There is no need to migrate your entire stack: our AI layer complements it rather than replacing it.

Kluwer
Navigator + Smartlaw
Sdu
Legal content
LexisNexis
Case law database
Clio
Case management
Filevine
Litigation management
Ironclad
Contract lifecycle
iManage / NetDocs
Document management
Own DMS / ERP
Custom integration

The AI layer sits alongside, not in front.

Kluwer, Lexis and Sdu do the heavy lifting on case law and legislation databases, and you don't need to replace them. We build an AI layer that unlocks your firm's own precedents, previous pleadings and internal playbooks through Retrieval-Augmented Generation. The model's suggestions always come with a source reference to your own material or to the legislation database, so there are no invented citations.

For clients with their own practice management system or legacy middleware, we build integrations case by case. That adds some design time to the planning, but prevents integration debt later. For deeper LLM integrations, see our page on custom LLM integrations.

Not sure about a large project yet?

Test your idea first: a working prototype in 1 day

With OneDayBuild, we make your idea tangible in a single day for €1,150, so you know whether further development is worth the investment. Decide to go ahead with the full build? Then we deduct the cost in full.

View OneDayBuild →

Which models and which use cases.

The choice of language model is not a cosmetic detail: it determines whether the solution fits your type of work. For long-context files, such as a 500-page deed or a complete case file, we generally use Claude because of its 200K-token context: the model reads an entire file in one go and delivers a structured summary without us having to chunk it artificially. For broad contract review flows with fast iteration we use GPT-4o, which is somewhat faster on shorter documents. And for firms with the strictest privacy requirements, such as trust offices, boutique M&A firms with government clients, or lawyers handling state-secrets cases, we deploy Llama 3.1 fully on-premise or in a private VPC. No outbound API connections, auditable model weights.

The use cases we build most often, grouped: document review and clause extraction for due diligence, RFP analysis and contract renewal cycles. Case file summarisation for litigation teams taking over a file or preparing a pleading. Semantic search across case law: not just keyword search on rechtspraak.nl or EUR-Lex, but conceptual search, such as "rulings in which a renewal clause in a lease was interpreted in the tenant's favour under ROZ conditions". Client enquiry triage via a chatbot front end on your website that handles first-line questions and gives the lawyer a structured summary ahead of the intake conversation. Legislation monitoring that scans the Staatscourant, the EU Official Journal and case law for relevance per client or file, with no more manual alerts.

We also build compliance mapping for client portfolios (GDPR, Wwft, AI Act, CSRD), KYC/Wwft screening for client onboarding, and, for firms that themselves advise clients on AI compliance, AI bias audit tooling that the firm uses to assess clients' applications. Time-recording automation is an underrated win: an AI that categorises email traffic, calendar entries and document edits against open files significantly reduces the administrative burden per lawyer, especially in firms under billable-hours pressure.

Whatever the use case, the approach is the same: a RAG layer that unlocks your own knowledge base, an EU-residency deployment of the model, an audit log that stores every model call with its input and output for review, and a human-in-the-loop design in which a lawyer always has the final check before output goes to a client.

From workflow audit to production in clear steps.

An AI implementation at a law firm has its own rhythm. We go through five phases in the same order for every firm.

01 · Audit

Use case selection

One day at your firm with partners and paralegals. Outcome: which flows deliver value immediately, which need more preparation, and which drop out on compliance grounds.

02 · Design

Architecture & model selection

Choose between Claude for long-context case files, GPT-4o for broad tasks, or Llama for private deployment. We map out the RAG strategy for your knowledge base and DMS.

03 · Build

Sprints with legal review

One working flow delivered every sprint. A dedicated partner checks the output for legal accuracy. No flow goes live without a lawyer signing off on its quality.

04 · Cutover

Phased rollout by team

One practice group first, then the next. With each rollout comes training on spotting hallucinations, prompt discipline and the NOvA limits on AI use.

05 · Maintenance

Ongoing, with an audit cycle

We monitor model drift, update with every new LLM version and keep the AI Act documentation current. Compliance is not a project but an ongoing obligation.

Context within the legal tech market.

Wolters Kluwer Future Ready Lawyer 2024

"79% of Dutch lawyers named AI as the technology with the greatest impact on the profession over the next three years — a figure that has doubled in a year."

NOvA, Guidelines on the use of AI 2024

"A lawyer remains ultimately responsible for every work product created with AI; the application must be set up so that accuracy and source references can be verified at any time."

European Lawyer's Tech Outlook 2025

"Mid-tier firms that are building custom AI on their own precedent bank now are noticeably ahead of firms waiting for commodity suites; the difference lies in the firm's own knowledge base, not in the model."

Answers for legal practices considering AI.

The questions we hear most often from managing partners, IT leads and compliance officers.

What does AI actually do in a law firm?
The most visible applications are document review on large batches of contracts, case-file summarisation for bulky litigation files, semantic search across your own precedents and case law, and first drafts of standard pleadings or advisory memos. Beyond that: automating KYC/Wwft screening, legislative monitoring per client portfolio, and client-query triage via chatbot, where the lawyer receives a structured summary. In every case the lawyer remains ultimately responsible — the AI does preparatory work, not autonomous decision-making.
Will our case file data end up with OpenAI or Anthropic?
No — that is one of the first design choices we make with you. We deploy LLMs in a private EU environment (Azure OpenAI within your own tenant, AWS Bedrock in an EU region, or a fully on-premise Llama stack). No training on your data, no retention by the model provider, no US Cloud Act risks. Your duty of confidentiality demands this, and we design around it — not the other way round.
How does this relate to the EU AI Act and the duty of confidentiality?
Legal AI falls under Annex III of the AI Act and is therefore high-risk. That means: documentation of the risk classification, human-in-the-loop design, data quality monitoring, and transparency towards the end user. We provide that documentation as standard. For the duty of confidentiality (Wbo art. 11a) we design data flows so that privileged information is not shared inadvertently. Firms that advise their own clients on AI compliance can use the same infrastructure — see our page on enterprise AI implementation.
How do you stop the AI from inventing case law?
Hallucination is the biggest reputational risk for legal AI — US judges have already sanctioned lawyers who cited fictitious rulings. We design every flow with mandatory source references: the model never cites a ruling without a verifiable link to rechtspraak.nl, EUR-Lex or your own precedent files. When in doubt, the model declines to answer and escalates the question to a lawyer. A RAG architecture with strict source grounding is essential here — we do not allow free LLM output without a source in production.
Does AI replace the lawyer?
No, and that is not a design goal either. AI is good at the heavy lifting — searching hundreds of contracts, summarising a 500-page file, drafting a first version of a memo. But legal judgement, strategic advice to the client and ultimate responsibility for every work product stay with the lawyer. The firms that get the best results use AI to free up paralegals and lawyers for high-value work — not to replace them.
Do we need to replace Kluwer or Lexis first?
Almost never. Kluwer Navigator, Sdu Juridisch and LexisNexis are strong in legislation databases and case-law publication — you don't need to redo that. The added value of custom AI lies in the combination with your own firm precedents, your own playbooks and your own files — that is where we build a RAG layer on top. We work happily alongside them and, where possible, integrate via the publishers' APIs.
What determines the cost of an AI project?
Three factors weigh most heavily: the scope (a single use case such as contract review versus a firm-wide platform with multiple flows), the compliance requirements (a private LLM deployment is more involved than a SaaS layer) and the depth of integration with your existing DMS and case management. We work with a fixed sprint budget and give a concrete price for the full build after the design phase — no open-ended contracts. For cost context on AI projects more broadly, see our page on enterprise AI implementation.
How do we get started with AI as a mid-size firm?
The recommended route is to choose one well-defined use case where your firm feels the pain — often contract review, file summaries or an internal knowledge base search function. Build a pilot around it with a partner as sponsor and a paralegal who co-develops. After a few sprints you have a working tool and — just as importantly — a team that understands what AI can and cannot do. Only then do you expand to other flows. For broader AI skills within the firm, we also provide AI literacy training for partners and staff.

Ready to bring AI into your legal practice?

A half-hour introduction with a managing partner or IT lead. We listen to your case flow, ask critical questions about compliance and give an initial direction. No obligation, and under strict confidentiality.

Edit content