7 Power Prompts for the Project Management

Structured Ideation With Large Language Models (LLMs)

(problem-solving, analysis, brainstorming, prototyping, decision support)

Purpose: Identify the systemic cause/ driver of a problem, bottleneck or failure through iterative causal questioning.
Prompt: “Act as a Consultant.
Here is the [CONTEXT].

Analyse [ISSUE] by
repeatedly asking 'why' until the root cause emerges (minimum five levels), then
propose one corrective action.”
Technique:
Continue beyond five if needed, then
challenge assumptions with “Why not?” and
drive ownership with “Why not us?”.

Purpose: Rebuild a strategy from fundamental truths, bypassing legacy constraints.
Prompt: "Act as a Business Consultant.
Apply first-principles thinking to [CHALLENGE]:
list all assumptions,
strip each to its fundamental truth,
propose a solution ignoring industry convention."
Technique:
Strip away all assumptions to solve the problem at its most essential level, then
rebuild the solution from scratch.

Purpose: Simulate a workshop, surface functional tensions before alignment by inhabiting three expert roles simultaneously
Prompt: “Take on three roles simultaneously ([Marketing Leader], [Sales Director], and [Product Head]) and critique [STRATEGY] from each perspective.

Then, for [TOPIC], complete this sequence:
A. Generate one bold idea per role (3 total);
B. Cross-critique each idea through the other two lenses: [Brand], [Revenue], [Feasibility];
C. Synthesise the strongest idea(s) into a single recommendation that
directly resolves the top objection raised.”
Technique:
Each expert produces ideas independently and the model reveals hidden dependencies, synergies, and synthesises common themes and trade-offs.
Use a chain of prompts instead of a sequence of actions.

Purpose: Generate innovations by modifying an existing product, process, or service through feature mutations.
Prompt: “Act as a Business Consultant.
Apply the SCAMPER framework to [PROCESS/ PRODUCT].

For each lens (Substitute, Combine, Adapt, Modify, Put to other use, Eliminate, Reverse),
generate one concrete innovation idea.

Format as a table.”
Technique:
Produce ideas using structured transformation

Purpose: Prevent failure by imagining the worst-case outcome(s) before commitment.
Prompt: "You are a senior risk advisor tasked with killing [PLAN/ PROJECT].

Identify the top 5 fatal flaws:
- strategic,
- financial,
- operational,
- market,
- execution.

Then propose a mitigation for each - prioritise by impact."
Technique:
Reverse the goal to uncover overlooked vulnerabilities, operational blockers and expose the weakest assumptions.

Purpose: Conduct a 360-degree review of a strategic decision to mitigate biases.
Prompt: "Act as a Consultant.
Evaluate [PROPOSAL/ DECISION] using Six Thinking Hats.
One focused laconic paragraph per hat:
- White=data,
- Red=gut feelings,
- Black=risks,
- Yellow=benefits,
- Green=creative alternatives,
- Blue=next steps & control."
Technique:
Force six cognitive modes to eliminate groupthink before any major decision.

Alternative: Scenario Modelling (Bull/Bear/Base)

Purpose: Maximise signal-to-noise ratio by subjecting ideas to a rigorous critical review process.
Prompt: “Act as a Business Consultant,
rank these [SOLUTIONS/IDEAS] from best to worst and explain WHY.
No hedging”
(Optional: + “use the following criteria: strategic impact, feasibility, and speed to value; score confidence on every claim”).
Technique:
Learn which LLM (ChatGPT, Google Gemini, Claude - each model has different training biases) works best as a Critic/ Editor or as Actor/ Solution Generator – for executive-ready recommendations.
Use disagreement between models as a signal!

Alternative: Chain of Density – if it converges to something meaningful

Iterate: Generate → refine → test → iterate.
Human-in-the-Loop: LLMs are probabilistic; they are hallucination-prone and over-confident. You are the judge; the LLM is a tool.
A good prompt includes 4 elements (if you drop any of them – it often degrades genAI output): Context, Role/ Persona, Task, Target Audience / Output expectation