DR-101 · Module 2

Prompting for Depth

4 min read

The default AI response to most questions is a surface-level overview — competent, accurate enough, and completely useless for actual research. The model gives you breadth because breadth is safe. Depth requires you to explicitly ask for it, and the way you ask determines how deep you actually get.

The role instruction is the most reliable depth technique. Instead of asking a generic question, assign the model a specific expert role with a specific analytical framework. "You are a competitive intelligence analyst with 15 years of experience in enterprise SaaS. Analyze Company X's go-to-market strategy using Porter's Five Forces" produces fundamentally different output than "Tell me about Company X's strategy." The role activates domain-specific vocabulary, analytical patterns, and depth expectations that a generic prompt does not.

  1. Assign a Role Define who the model is for this conversation: "You are a [specific expert] with [specific experience]." The role sets the depth expectation and analytical vocabulary.
  2. Specify the Framework Name the analytical lens: SWOT, Porter's Five Forces, Jobs-to-be-Done, First Principles. A framework forces structured analysis instead of freeform commentary.
  3. Set the Audience Define who will read the output: "Write this for a VP of Sales who needs to make a decision by Friday." Audience determines depth, jargon level, and actionability.
  4. Demand Evidence Add: "Support each claim with specific evidence. Flag any claims where you are less than 80% confident." This forces the model to self-audit and reduces confident hallucination.

The framing technique adds another layer. Instead of asking what something is, ask what it means. Instead of asking for a description, ask for an implication. "What are the implications of Company X's pivot to vertical SaaS for mid-market buyers in the manufacturing sector?" forces the model to analyze cause and effect, not just report facts. Implications, trade-offs, second-order effects — these framing words push the model past the surface layer.