The Financial AI Prompt Library

• Audit support and audit programmes
• Financial analysis and reporting
• Management letters and long-form audit reports to those charged with governance
• Questions on regulation and standards
• Client communication
• Materiality and risk assessment
• Confidentiality and safe data entry
• Quality management, CPD and AI policy
• The Assurance Safety Framework, with five diagrams and supporting illustrations
• Practical details: one hundred fill-in prompts across eight categories, approximately 100 pages, English edition 2026, delivered as PDF and as Word so that you can copy, complete and adapt the prompts to your own firm's house style.

The Financial AI Prompt Library is a workbook, not an essay about artificial intelligence. It contains one hundred ready-made, fill-in prompts that you complete with your own case, organised around the work you actually do, from the audit programme to the management letter, from financial analysis to your own firm policy.

A language model predicts the most probable text, and an invented amount looks exactly like a real one. Without a fixed framework that forces every figure and every reference back to its source, the model produces convincing fiction. This library gives you that framework. Every prompt that asks for a figure, a calculation or a standard forces the output to be traceable and marks everything not yet checked as [TO VERIFY].

That is how the speed of the machine stays within the boundaries of your professional rules, and how judgement stays where it belongs, with you.

• One hundred fill-in prompts across eight categories: audit support and audit programmes, financial analysis and reporting, management letters and long-form audit reports to those charged with governance, questions on regulation and standards, client communication, materiality and risk assessment, confidentiality and safe data entry, and quality management, CPD and AI policy.
• Traceability and reproducibility as the centrepiece: no amount, no calculation and no citation leaves your file without substantiation you can follow, aligned with the building blocks the AFM (Dutch financial markets authority) sets for responsible use of technology in the audit.
• The anti-hallucination protocol built in: the model may not invent an amount, a figure, a standard or a source, shows every calculation step by step, and marks every assumption as [TO VERIFY] until you have checked it yourself.
• A purpose, a completion instruction and a safety note with every prompt, so that you know what the prompt delivers, how to complete it and where you have to look yourself.
• Materiality and professional scepticism safeguarded: the model organises and drafts, but weighs no materiality and draws no conclusion, that remains your judgement.
• Confidentiality and data security safeguarded: no client or audit data in public tools, with final responsibility resting with you and every step recorded in an audit trail, aligned with the VGBA (Dutch code of ethics for accountants) and the GDPR.
• The Assurance Safety Framework, with five diagrams and supporting illustrations, translating the VGBA principles, the auditing standards and the EU AI Act into practical working arrangements.
• Grounded in real sources and real cases: the building blocks of the AFM, the guidance of the NBA (Dutch institute of chartered accountants) and NOREA (Dutch association of IT auditors), the VGBA and its requirement of professional scepticism, the auditing standards, and the internationally discussed case of a government report containing sources invented by AI.

For the engagement partner who signs the opinion and personally stands behind it. For the controller and the finance professional who substantiates the figures the organisation steers by, and for the owner or partner of an accountancy firm who guards the quality of the output. And for the sole practitioner who works without a second pair of eyes.

In short, for the person who, once the signature is given, is answerable under professional disciplinary law, under civil law and in reputational terms. The model does not carry that risk. You do.

There is no shortage of free prompt lists in circulation. What they lack is precisely what a financial practice needs, namely a framework that makes the output safe. Here that framework is the core, not an appendix.

We do not sell hype and we promise no miracles. We separate fact from opinion, refer only to real sources named in the text, and build the safeguards into the prompt rather than into a disclaimer afterwards. That is necessary, because the ground is shifting. The use of AI in the audit and in the finance function is increasing sharply, the AFM (Dutch financial markets authority) and the NBA (Dutch institute of chartered accountants) have sharpened their expectations, and the EU AI Act imposes requirements. That is the field in which you sign.

And because we know where things go wrong, we have held ourselves to our own rule. This product contains no invented amount, no non-existent standard and no fabricated source. What we teach you to demand of the machine, we first demanded of ourselves.

Browse the eight categories and the verification protocol first if you wish. You do not have to decide anything yet. You judge for yourself whether it fits the way you sign.

The library is built on real, named sources: the building blocks of the AFM (Dutch financial markets authority) for responsible use of technology in the audit, the guidance issued by the NBA (Dutch institute of chartered accountants) and NOREA (Dutch association of IT auditors), the VGBA (Dutch code of ethics for accountants) with its requirement of professional scepticism, the auditing standards, the GDPR and the EU AI Act, together with the internationally discussed case of a government report containing sources invented by AI. No amount, standard or source in this product is fabricated. Where a point still requires checking against your own situation, the text says so and marks it as [TO VERIFY].