Before you automate a single HR task, you need a reliable HR co-pilot — a general-purpose assistant (Claude, ChatGPT, Microsoft Copilot, or Gemini) configured with the context, tone, and guardrails of your HR function. A co-pilot with no setup gives generic answers. A co-pilot with a good system brief becomes a junior HR business partner who never sleeps.

Why a system brief matters

Large language models respond to the frame you give them. If you tell the model who it is, who it serves, what data it may and may not touch, and what ‘good’ looks like, every later answer inherits that frame. In MENA HR you also need it to respect local context: Arabic name conventions, Hijri and Gregorian dates, Ramadan working hours, and labour-law differences between Egypt, KSA, and the UAE.

Step-by-step: build your co-pilot brief

  1. Open a fresh chat in your tool of choice. In Claude or ChatGPT, save this as a Project / Custom Instruction so it persists.
  2. Define the role, the company, the data rules, and the output style.
  3. State explicitly what the model must refuse — e.g. making a final hire/fire decision, or processing real employee PII.
  4. Paste in 3–5 anonymised facts about your org (headcount, functions, locations, languages).
  5. Test it with one easy question and one edge case before you trust it.

Use this starter brief and adapt the bracketed parts:

You are my HR co-pilot for [Company], a [sector] firm with [N] staff across [Cairo / Riyadh / Dubai].
Working languages: English and Arabic; default to English unless I write in Arabic.
Jurisdiction: primary labour law is [Egyptian Labour Law / Saudi Labor Law]; flag when a question is jurisdiction-sensitive.
Data rule: I will only ever paste SYNTHETIC or anonymised data. Never ask me for real employee national IDs, salaries tied to names, or health data.
Always: give me a recommendation first, then the reasoning, then 2 risks.
Refuse: do not make a final hiring, promotion, or termination decision for me — produce options and a recommendation only.
Confirm you understand by listing the four rules back to me in one line.

Try it

Set up the brief above in your preferred tool, then ask: “Draft a one-paragraph internal announcement that we are piloting AI assistance in HR, reassuring staff their personal data is protected.” Check that the tone matches your culture and that it did not invent any policy you do not have. Save the working brief — you will reuse it in every later lesson.

Next