Free to copy. No form, no e-mail address.
This is a one-page AI policy an accounting, administration or tax firm can start from: which tools staff may use, which client data may go in, and who checks what goes out. Copy it, change it, and use it with or without Bee Flow.
Plan a callVersion 1. Fill in the name of your firm, the date, the approved tools and the AI contact.
This is a starting text and not legal advice. Have the person who handles privacy in your firm read it before you adopt it.
We use AI to prepare our work: drafts, summaries, explanations and first checks. This policy says how we do that with client data. It applies to everyone who works for our firm, including temporary staff and trainees.
We use AI for work only through the tools on our approved list, with an account that the firm manages. Personal and free accounts are not used for work. For each tool the list says whether client data may go in. A tool is suitable for client data only when it is a closed environment, the firm has a processing agreement for it, and we know where the data goes and who can reach it. Want to use a tool that is not on the list? Ask the AI contact first.
Client and personal data goes only into a tool that the list marks as suitable for client data. It never goes into a public or free AI tool. Even in an approved tool you enter no more than the task needs. Leave out names, the BSN, account numbers and addresses where the question works without them, or describe the case without them. Passwords, access codes and login details never go into any AI tool.
AI prepares, a colleague decides. Whoever sends or signs something answers for it. Before a text, a calculation or an advice goes to a client or an authority, you read all of it and you check amounts, dates, names, references to law and rules, and everything we promise. An AI answer alone is never the basis for a conclusion in a client file. You check it against the source documents.
Where AI contributed to work in a client file, you note in the file: which tool you used, what you asked (the prompt or the search terms), which sources were consulted, which checks you did and what you concluded. Prompts that the firm uses again and again are kept in one shared place, with the name of the colleague who checked them.
Questions go to the AI contact. That is one named colleague. Did you enter something that should not have gone in? Tell the AI contact the same day. Nobody is blamed for reporting. Together with the person who handles privacy, the AI contact decides whether it is a data breach that has to be reported.
We review this policy after every busy season, and whenever a tool is added or dropped. The version and the date are at the top, so everyone can see which text applies.
For firms that follow the NBA and NOREA guide
What AI-leidraad 2: AI Toegepast (NBA and NOREA, June 2026, sections 2.1 and 2.2) asks, next to the Bee Flow setting that helps with it and the licence that setting is in. The quotes are from the Dutch original.
The quotes were checked against the published PDF on 16 November 2026. This table shows which setting helps with which condition. It is no statement that a firm using Bee Flow meets the guide, and no software makes a firm meet it.
| What the guide asks | The setting, and its licence | |
|---|---|---|
| A closed environment | “Deze data moet in een beveiligde, afgeschermde omgeving blijven.” (2.1) | Your own server or Bee Flow Cloud in the EU, on every licence. One switch limits every model level to EU-hosted models, and a model on your own hardware keeps text on your network. |
| Client data and public models | “Cliënt- en persoonsdata mag niet zomaar in publieke AI-modellen worden ingevoerd” (2.1) | The personal-data filter detects and blocks 21 kinds of personal data (Free, guard service installed) and replaces them before the model sees them (Pro). Detection can miss. |
| Responsibility | The professional keeps “de volledige en ongedeelde verantwoordelijkheid” (2.1) | An approval step with a record of who decided, when, why and what they saw (Pro, where a step is built in). Rights per app and per action: read, ask first, always ask (Free). |
| No black box | “Daarbij is black box-AI niet te verantwoorden en een onacceptabel kwaliteitsrisico.” (2.1) | Under an answer: the tools and documents used and what the filter did (Free). Every automation run kept step by step (Free). |
| Sources in the outcome | “In de uitkomsten van de AI-toepassing verwijzingen opnemen naar de gebruikte bronnen.” (2.2, point 1) | Answers from a knowledge base cite the document, page or row (Free). A plain chat answer has no source list. |
| A record in the file | “Het vastleggen in het dossier van gehanteerde prompts, zoektermen, geraadpleegde bronnen, de signalen, uitgevoerde verificatiestappen en de uiteindelijke conclusies” (2.2, point 7) | The trace under an answer and every run step by step stay available in Bee Flow (Free). Outgoing tool and automation calls are logged without content, with the report in Advanced monitoring (Pro). Putting it in your file is your own step. |
| A second AI as a check | “Dit kan bijvoorbeeld op basis van een tool of LLM van een andere leverancier.” (2.2, point 8) | Nine hosted model providers and models on your own hardware. You can put the same question to a model from another provider. |
| AI output is not evidence | “Alleen AI-output kan nooit als voldoende en geschikte controle-informatie worden beschouwd.” (2.1) | No setting replaces this. Bee Flow prepares, and your professional judges. |
Three steps that fit in one team meeting.
Ask first who uses which AI tool and for what, without judging the answers. Then go through the seven points and show one example with the client details left out.
One colleague is the AI contact: for questions, for new tools and for the message "I pasted something I should not have". Put the name on the page.
Put a date on the page and a fixed item on the agenda: are we using anything new, and did anything go wrong.
There is no download and no form. Print this page, or copy the seven points into your own template and fill in the name of your firm, the approved tools and the AI contact. The guide itself is a PDF on the NBA website, in Dutch.
Read AI-leidraad 2 on nba.nl (PDF, Dutch)
Forward this page. The facts they will ask for are written down, including what is missing.
Prompts, files and answers: who can reach them, which model providers are involved and what is logged.
Sign-in, roles, retention, encryption, and what is not covered.
One command with Docker on amd64, or Bee Flow Cloud in the EU.
FAQ
Yes. Copy it, shorten it, add your own rules and put your own name on it. You do not have to mention Bee Flow.
No. A policy is one part. Your firm still owns the processing agreements with its suppliers, and the DPIA where one is required. This text is not legal advice.
No. The policy holds with any tool that meets point 2. The table on this page shows where Bee Flow helps, for firms that want one environment for it.
If your firm has a works council, it may. Under article 27 of the Dutch Works Councils Act (WOR), the works council has a right of consent on rules about systems that can be used to monitor the attendance, behaviour or performance of staff. Whether that applies to the AI tools you choose depends on how they are set up, so discuss it with your works council before you introduce them. The article on AI and the works council under Read more goes into this.
The explanation behind the policy.
What goes on the page, how to introduce it in one meeting and how to keep it current.
When you need consent, and what to bring to the meeting.
What happens to client data in Bee Flow, and what we do not claim.
20 minutes. An honest go or no-go, and what a fixed price would look like.
Plan a call For accounting and tax firms