AI Tip's PubAI Tip's Pub
Back
ChatEND USER5 min read

Run a Deep Research investigation

On a complex topic, let Copilot “think” longer: pick the right reasoning mode (Quick response vs Think deeper) to get a sourced, actionable synthesis.

How to

  1. 1What is deep research? It's not a simple Q&A: you ask Copilot to explore a topic from several angles, cross-check sources, compare options and return a structured synthesis (context, facts, assumptions, recommendation). Great for a decision, a benchmark, a scoping note or market watch.
  2. 2Choosing reasoning power: the mode selector. In the top-left of Copilot Chat, a selector (“Auto” by default) decides how long Copilot “thinks” before answering. This setting is what separates an instant reply from a genuine in-depth analysis.
    Run a Deep Research investigation : étape 2
  3. 3Understand the 3 modes: ➞ Auto: Copilot decides how long to think based on the question. The sensible everyday default. ➞ Quick response (GPT-5.5): answers right away, with no long reasoning step. Built for speed. ➞ Think deeper (GPT-5.6): Copilot thinks longer to build a structured, more reliable line of reasoning. Built for quality on complex topics.
  4. 4Which mode for which request? ➞ Use Quick response for: a factual question, a definition, a rewrite, a short email, quick brainstorming, when speed matters and stakes are low. ➞ Use Think deeper (or Auto) for: analysis, multi-criteria comparison, strategy, synthesis across several sources, multi-step reasoning, when quality matters and a few seconds' wait is fine. Simple rule: low stakes + need speed → Quick response; high stakes + need reliability → Think deeper.
  5. 5Launch the deep research. Select Think deeper (via GPT ➞ “GPT 5.6 Think deeper”), then state a precise question AND the expected deliverable (brief, comparison table, reasoned recommendation).
  6. 6Set the quality bar: ask Copilot to cite its sources and clearly separate facts from assumptions.
  7. 7Iterate to dig deeper: “Go deeper on point 2”, “Compare the 3 options in a table”, “What are the risks and blind spots?”.

Prompt to copy

Run a deep research on [topic] (Think deeper mode). Deliverable: structured brief with context, 3 compared options (table), risks, and a reasoned recommendation. Cite your sources and separate facts from assumptions.
#recherche#deep research#think deeper#gpt-5.6#analyse

💬 Comments

No comments yet. Be the first!

Discover more