1.3. The Copilot Mental Model: Assistant, Not Automaton

💡 First Principle: Copilot amplifies human judgment — it does not replace it. The most effective use of Copilot follows a loop: AI generates a draft → human reviews and refines → human takes responsibility for the final output. Breaking this loop is where professional and ethical risks emerge.

A useful mental model: think of Copilot as an extremely capable new colleague who is always enthusiastic, always available, and always confident — but who you should never let submit work without review. That colleague can dramatically accelerate your output. But they can also confidently send an email with the wrong figures to a client if you are not paying attention.

Three practical implications of this mental model:

1. You are always the author. When Copilot drafts an email and you send it, you sent that email — not Copilot. The professional accountability stays with you.
2. Copilot is best at acceleration, not origination. Starting with an AI draft and refining it is almost always faster than writing from scratch. But completely delegating a high-stakes decision to AI without review is where things go wrong.
3. The quality of your input determines the quality of the output. Vague prompts produce vague results. Specific, well-structured prompts produce specific, useful outputs. You will spend time on the prompt or you will spend time fixing the output — investing in the prompt is almost always more efficient.

⚠️ Exam Trap: The exam will test your understanding of "over-reliance" — scenarios where professionals trust AI outputs without sufficient verification. The right answer in these scenarios always involves human review steps, especially for sensitive, high-stakes, or externally-facing content.

Reflection Question: A manager asks Copilot to summarize a 50-page contract and plans to rely on the summary for a negotiation meeting without reading the original. What risk does this represent, and what should the manager do instead?