Legal Prompts That Actually Work

• The discipline-safe framework: the lawyer remains responsible, every output is verified, client data never goes into a public AI tool
• The anatomy of a working prompt: role, context, task, format and safeguard
• Category 1: Contract analysis and drafting contracts
• Category 2: Litigation documents and legal drafts
• Category 3: Case-law and literature research
• Category 4: Client communication and letters
• Category 5: Due diligence and compliance
• Category 6: Office productivity
• Category 7: Negotiation and strategy
• Category 8: Legal knowledge, memos and explanation
• Cases from practice, including Mata v. Avianca and Dutch cases
• 5 diagrams on prompt structure and the workings of the framework

This is not a book about the promise that a language model will take over your work. It is about the opposite. You remain the lawyer, you keep the responsibility, and you learn to direct the model so that it saves you time instead of creating risk.

The difference lies in how the question is built. A prompt that sets out role, context, task, format and safeguard turns a blank screen into a usable first draft. Not the finished product, but the starting point you would otherwise have written from nothing.

The prompts are the packaging. The discipline-safe framework, meaning safe under the professional disciplinary rules that govern Dutch lawyers, is the actual product. This library hands you that working method one hundred times over, ready to use, situation by situation, each with a safety note telling you exactly where to look twice.

• 100 ready-to-use, fill-in prompts across 8 categories: contract analysis and drafting contracts, litigation documents and legal drafts, case-law and literature research, client communication and letters, due diligence and compliance, office productivity, negotiation and strategy, and legal knowledge, memos and explanation.
• Every prompt comes with a purpose, guidance on how to fill it in and a safety note, so you know what you are asking, how to complete it and where verification is required.
• A discipline-safe framework at the heart of the book, built on three rules. The lawyer remains responsible, every output is verified, and client data never goes into a public AI tool.
• Grounding in real sources, including the rules of conduct and core values of the Dutch Bar (gedragsregels en kernwaarden), the GDPR and the EU AI Act, with fact and opinion kept properly apart.
• Instructive cases from practice, including the American case Mata v. Avianca and Dutch cases, so you can see what goes wrong without proper instruction and how to prevent it.
• 5 diagrams and imagery that make the anatomy of a good prompt and the workings of the framework visible.
• Immediately usable, 98 pages, in English, 2026 edition, in PDF and Word so you can copy and adapt the prompts.

For the lawyer in solo practice who wants to gain time without exposure under professional disciplinary law. For the in-house counsel at an SME who runs many files with a small team and needs structure. And for the paralegal who wants to deliver the groundwork faster and more cleanly, with a result that can be checked.

It is written for those who have already tried AI and been disappointed, and for those still hesitating because the risks are unclear. In both cases you receive a working method here rather than a promise.

BNK KAM is not selling you a miracle. We choose restraint deliberately. No tenfold acceleration, no suggestion that the machine takes over the thinking. What sets this book apart is that the safety is sold with it rather than polished away.

Fact. In the American case Mata v. Avianca in 2023, lawyers were fined because they relied on case law invented by ChatGPT. Fact. In the Netherlands too, careless use has had consequences. Lawyers received a warning and a mandatory AI course, a court set aside the case law cited by a lawyer who twice came forward with invented rulings, and the Raad van Discipline (Dutch disciplinary board for the Bar) in ’s-Hertogenbosch issued a reprimand to a lawyer who had adopted AI-generated citations without verification. A suspension has occurred in this domain abroad, not in the Netherlands. That is precisely what we report, and nothing more.

These examples are not in the book to dramatise, but to show what exactly goes wrong, and how simply it can be prevented. The framework rests on real foundations. The rules of conduct and core values of the Dutch Bar, including the duty of confidentiality. The GDPR for the protection of personal data. The EU AI Act for the wider framework within which this technology is developing.

Opinion. We are convinced that most lawyers ask too little of AI rather than too much. What makes the difference is not more output, but better instruction and a fixed safety ritual. That is why the discipline-safe framework is not an appendix here, it is the actual product. Where we make a recommendation, we say that it is a recommendation. Responsibility for the final result rests with you.

If you want to see what a well-built prompt delivers, start today with the category that takes the most work off your hands.

The book cites what can be verified and says so plainly. The American case Mata v. Avianca from 2023, in which lawyers were fined for relying on case law invented by ChatGPT, and Dutch cases in which careless use led to a warning or a required course, including one lawyer who twice submitted invented rulings to the same court. A suspension has occurred in this domain abroad, not in the Netherlands, and that is how the book puts it. The framework itself is grounded in the rules of conduct and core values of the Dutch Bar (gedragsregels en kernwaarden), including the duty of confidentiality, the GDPR for the protection of personal data, and the EU AI Act for the wider regulatory picture. Fact and opinion are separated throughout, and where a statement is a recommendation, it is labelled as one.