How to Use AI to Write a Text: A Practical Guide
To use AI to write a text, follow four steps: define precisely what you want (the prompt), choose the right tool for your need, review and correct the generated output, then adapt the tone to your context. AI does not replace your judgment. It accelerates the production of a first draft that you then finalize.
The Real Problem: You Lack Time, Not Ideas
The HR director who needs to write a job posting, a rejection email, or an internal policy memo isn’t short on ideas. They’re short on time. And often, they’re stuck on the first sentence.
That’s exactly where AI is useful. Not to think for you. To start for you.
This guide is written for you if you’ve never used an AI writing tool, or if you tried it once and got something unusable.
Step 1: Choose the Right Tool
You don’t need a list of twenty tools. You need a starting point.
For professional writing in French, three tools cover most needs:
- ChatGPT (OpenAI): the most versatile. Writes emails, reports, summaries, job postings. Available in free and paid versions.
- Claude (Anthropic): very strong on long texts and tone nuance. Useful for executive memos or sensitive HR documents.
- Mistral (French): a sovereign option if data confidentiality is a regulatory constraint in your organization.
If you’re starting out, begin with ChatGPT. The learning curve is short and results are immediate.
To compare these tools in detail, I published a comparison of the best free AI tools in 2026 that will give you an overview before you choose.
Step 2: Build a Prompt That Works
This is where most people fail. They write “write me an email” and get something generic and unusable.
An effective prompt contains four elements:
- The role you give the AI (“You are a senior HR director at a services company”)
- The precise task (“Write a post-interview rejection email”)
- The context (“The candidate was qualified but we chose someone internally”)
- The expected tone (“Professional, warm, short, under 150 words”)
With these four elements, you get a usable first draft in thirty seconds.
Without them, you get generic text that will take longer to fix than to write yourself.
Step 3: Iterate, Don’t Accept
The first output is never the final version. It’s a foundation.
You read it. You identify what doesn’t match your tone, your company culture, your relationship with the recipient. And you ask the AI to adjust.
“Make this paragraph more direct.” “Remove the excessive courtesy formulas.” “Add a sentence about our commitment to follow up with the candidate in six months.”
Each instruction refines the result. Two or three iterations are generally enough to arrive at something you can sign without hesitation.
This is what I observe among executives who integrate AI into their daily work: they don’t look for perfection on the first try. They treat AI like a junior collaborator you brief and correct.
I’ve built a methodological framework to structure this approach for leadership teams. Download the Board Pack AI 2026 to see how to apply it to your organization.
Step 4: Review With Your Own Judgment
AI can produce fluent texts that contain factual errors, legal approximations, or formulations that don’t match your reality.
An AI writing tool doesn’t know your collective agreement. It doesn’t know that your CEO hates emails longer than three paragraphs. It doesn’t know that your Moroccan client prefers a formal opening before getting to the point.
You know these things. That’s why reviewing is not optional.
The simple rule: if you can’t defend every sentence in that text, don’t send it.
Step 5: Build Your Own Templates
Once you have a prompt that works for a type of text, save it.
The executives and HR directors who genuinely save time with AI don’t reinvent their prompt every time. They have a library of tested prompts: job postings, rejection emails, executive memos, meeting minutes, interview summaries.
This is a team resource. Shareable. Collectively improvable.
This connects to the broader question of AI culture in organizations, which I address in my article on the role of AI in human resources management.
Pitfalls to Avoid
Three mistakes come up consistently among those who abandon the tool after two attempts.
First pitfall: the vague prompt. “Write me something about our HR policy” produces nothing useful. Be precise about format, length, recipient, and objective.
Second pitfall: sending the raw output. AI writes fast. You read fast. And you send a text that doesn’t sound like you. Your recipient feels it. Take two minutes to review.
Third pitfall: using AI for high-stakes texts without human validation. A contract, a crisis communication, a legal document. These require expertise that AI simulates but does not possess.
What You Can Concretely Expect
Executives who adopt this method consistently describe the same thing: they spend less time staring at a blank page and more time refining substance.
A job posting that used to take an hour now takes fifteen minutes. A difficult email to formulate comes out in five minutes instead of twenty. An executive memo structures itself in a few exchanges with the tool.
This isn’t magic. It’s method.
If you want to structure this approach for your team or organization, request a free diagnostic. We’ll look together at where AI can generate measurable value in your communication and management processes.
FAQ
Can AI write texts in Arabic or Darija?
ChatGPT and Claude handle standard Arabic with decent quality. Moroccan Darija remains limited. For internal communications in Arabic, plan for systematic native review.
Do you need to pay for a useful AI writing tool?
The free version of ChatGPT is sufficient for most everyday professional writing needs. Paid versions bring speed, file attachments, and more recent models. Start free, upgrade to paid if you use it daily.
Can AI replace a writer or an HR assistant?
No. It accelerates certain text production tasks. It does not replace the judgment, relationship, and contextual knowledge that a human collaborator brings. It’s a tool, not a position.
Is my data safe if I use ChatGPT for HR texts?
Never paste personally identifiable data into a public AI tool. No candidate names, no ID numbers, no medical information. Use fictional or anonymized data in your prompts.
How do I know if the generated text is good quality?
Ask yourself one question: would I sign this text as is? If the answer is no, keep iterating or rewrite the passages that don’t work for you. Final quality remains your responsibility.