Planning and Decisions

Prompting for Planning and Decisions

Planning and decisions are where AI's breadth genuinely helps: more options, faster drafts, blind spots surfaced. But only if you prompt for thinking rather than for answers. The patterns here are not "AI tools." They are established decision-quality techniques that AI makes faster to run.

Planning Prompts

Breadth and Structure First, Judgment Second

Options generation

"Generate four distinct approaches to [goal], each with a different tradeoff. Do not recommend one yet."

AI widens your option set before you commit. The "do not recommend" instruction keeps it in the breadth role.

Risk enumeration

"List the ways this plan could fail and which assumption would break first."

AI surfaces risks you missed under time pressure. You judge which are real and which are generic.

Summary and synthesis

"Summarize this long document and pull out the three decisions that still need to be made."

Spot-check the summary against the original, especially mid-document content. That verification habit is covered in depth in the planning module.

Decision Prompts

Force the Slow Thinking That Confidence Skips

These three patterns force the deliberate reasoning that confident AI output tends to short-circuit. Note that none of them asks the AI to decide. Each asks it to make your own thinking better.

Pattern 1

Devil's Advocate

"I am about to recommend [decision]. Argue the strongest possible case against it. Focus on risks specific to this situation, not generic concerns."

A deliberate System 2 override against your own confirmation bias. Most useful when you are already sure you are right.

Pattern 2

Pre-Mortem

"Assume this decision failed badly six months from now. List the most likely reasons why."

Prospective hindsight (assuming the failure already happened) increased accurate risk identification by roughly 30% in Gary Klein's research (HBR, 2007). AI makes this exercise fast enough to actually run.

Klein, "Performing a Project Premortem," HBR 2007

Pattern 3

Reframe

"Restate this problem from the perspective of [a different stakeholder, a customer, a skeptical executive, the affected team]. What looks different from there?"

Changes the frame without changing the facts. Often surfaces the concern you had been quietly avoiding.

Reza's story

Reza is about to recommend promoting a strong, well-liked individual contributor to a first-time team lead role. Instead of asking AI whether to do it, Reza prompts for the devil's advocate: "Argue the strongest case against this promotion, focused on risks specific to the IC-to-manager transition." The AI returns sharp, specific dissent: the risk that technical excellence does not predict management skill, the loss of a top contributor, the support this person would need that the team cannot currently give. None of it is a verdict. All of it makes Reza's decision better.

Apply It Today

Run the devil's advocate on one real decision

Take one real decision you are sitting on and run the devil's advocate prompt on it. Notice whether it surfaces a risk you were quietly avoiding.

Reflect

  • Which planning bottleneck (blank page, narrow options, unseen risks) costs you the most?
  • On your last big decision, would a pre-mortem have changed anything?