The Daily AI Workflow (Desktop Print-Out)
Keep this routine on your desk:
- Open LibreChat first before diving into heavy drafting.
- Select your model based on the task complexity (see the Model Guide).
- Label your prompt using structured brackets (
[ROLE],[CONTEXT]) — see Prompting Methods. - Critique the output using the Executive Checklist below.
- Iterate: Tell the model what it missed and ask for a revised draft.
The Executive Safety Checklist (What to Trust vs. Question)
As a public director, accountability stops with you. Never publish AI output without running it through these five mental checks:
1. Is the claim verifiable? Can you cross-reference a specific statistic with a trusted regional database (like the ASEAN Food Security Information System — AFSIS, FAO, or ministry portals)?
2. Is the output overly generic? If the AI's recommendations could apply to any trading bloc in the world, it hasn't factored in the unique geopolitical, infrastructural, or economic realities of ASEAN. Push back and say: "Make this specific to our regional maritime logistics and existing trade agreements."
3. Is it displaying "blind confidence"? AI models write with absolute authority even when they hallucinate a fact or misinterpret a policy line. If a data point looks surprising, ask it directly: "What exact line in the uploaded text supports this number?"
4. Did I provide enough background? If the AI's output feels basic, it's usually a reflection of a basic prompt. Feed it template examples of how your ministry prefers multilateral briefings to be framed.
5. Would I sign it? The AI is an efficiency tool, not a replacement author. If you wouldn't confidently defend that specific page or briefing in a high-level ministry or international summit meeting, keep editing.