AI Lab · Free tool
Business AI assistant designer: system instructions and test plan
Specify your chatbot (WhatsApp, website, Instagram) like a pro: role, knowledge, languages including Darija, forbidden topics, human handoff. Get the system instructions, an anti-manipulation test plan and the go-live checklist.
🔒 Everything is computed in your browser: nothing is sent to or stored on a server.
1. Your business
2. Mission and channel
3. Languages and style
4. What the assistant knows
Paste your real information: it is the only source of truth. Whatever is missing here, it will say it does not know.
5. Limits and safety
6. Handing over to a human
System instructions
Completeness
Test plan
Send each message to your assistant and compare its answer with the expected behaviour. Run the tests again after every change to the instructions.
| # | Category | Test message | Expected behaviour |
|---|
Go-live checklist
Good practice and sources
- Give a precise role, a clear scope and an answer format: the basis recommended by the OpenAI, Anthropic and Google guides.
- Clearly separate your instructions from the data customers provide, and tell the assistant that this data is never an order (prompt injection, risk number 1 in the OWASP 2025 list).
- Put no secret in the system instructions: they can leak (OWASP "System Prompt Leakage" risk).
- Give the assistant only the access it strictly needs: no refunds or order changes without human approval.
- Test before launching, then monitor: manipulation attempts evolve.
Legal framework
- European Union: since 2 August 2026, Article 50 of the AI Act (EU 2024/1689) requires informing people, at the latest at the first interaction, that they are interacting with an AI system, unless it is obvious.
- Morocco: conversations contain personal data (name, phone number, address); processing them falls under Law 09-08 and the CNDP. Inform your customers and limit retention.
- This tool helps you draft: it does not replace legal advice for your situation.
Limits
Well-written instructions greatly reduce mistakes, but no AI model is infallible: an assistant can still be wrong or be manipulated. Keep a human in the loop for anything that commits the business (prices, refunds, health, contracts).
Sources
- OpenAI — Prompt engineering
- OpenAI — Safety best practices
- Anthropic — System prompts
- Anthropic — Mitigate jailbreaks and prompt injections
- Google — Gemini system instructions
- OWASP — Top 10 for LLM Applications 2025
- Commission européenne — AI Act, article 50
Need an assistant connected to your store, calendar or WhatsApp Business? See the AI and automation service
Why this tool
One source of truth
The assistant only answers from the information you paste (opening hours, prices, delivery). If something is missing, it says so instead of making it up: the number one cause of chatbot mistakes.
Tested before going live
The test plan covers normal cases, off-topic questions, missing information, "ignore your instructions" attempts, sensitive data and unhappy customers.
Built for Morocco and many languages
Darija in Arabic or Latin script, Arabic, French, English: the assistant replies in the customer's language, and the tests check it.
Frequently asked questions
Which tool can I use these instructions with?
Any platform that accepts system instructions: the OpenAI, Anthropic or Google APIs, a custom GPT, a chatbot platform or an n8n workflow. Paste the text into the "instructions", "system prompt" or "role" field.
Can the chatbot book appointments or check my orders?
Only if it is connected to your tools (calendar, store) through an integration. Without one, the generated instructions make it collect the details and pass the request to a human, without promising anything.
Do I have to tell customers they are talking to an AI?
Yes, and the option is on by default. In the European Union, Article 50 of the AI Act has required it since 2 August 2026 for systems that interact with people, and everywhere it is a matter of trust.
Is my information sent anywhere?
No. The text is assembled in your browser. In any case, never paste secrets (passwords, API keys) into system instructions: they can leak.
Going further
Automate your business with n8n and AI: 5 concrete scenarios
Read the guide →Play Store listing generator
Open the tool →llms.txt and AI robots
Open the tool →Coding agent instructions file
Open the tool →Prompt library by profession
Open the tool →Automation ROI calculator
Open the tool →AI API cost calculator
Open the tool →Anonymiser before sending to an AI
Open the tool →Local AI hardware calculator
Open the tool →Company AI usage policy
Open the tool →Bring AI into your business, for real?
Automations, assistants, multilingual content, apps: I design solutions that are useful, measurable and respectful of your data.