Goal: Use AI like a power tool, not a toy. Ship better outcomes with less chaos.
- Delegation What it is: Deciding what to hand off to AI, what to keep, and what to co-drive. Best to delegate
Repetitive work: drafts, boilerplate, summaries, formatting
Exploration: ideas, options, tradeoffs, edge cases
Execution scaffolding: checklists, templates, plans, test cases
“First pass” thinking: outline → critique → refine loop
Do not fully delegate
Final accountability: decisions, approvals, commitments
Sensitive context: confidential data, legal/medical certainty
High-stakes nuance: negotiations, policy, ethics, reputation moves
Delegation prompts
“Do a first draft with 3 variants, then critique each, then propose a final.”
“Generate a plan + risks + rollback strategy + success criteria.”
“Turn this into a checklist with owner, timebox, and definition of done.”
- Discernment What it is: Judging quality. Detecting hallucinations. Knowing when AI is confidently wrong.
Discernment checklist
Source of truth: What is the ground reality I can verify?
Assumptions: What did the model assume that I did not say?
Failure modes: What could be subtly incorrect but sound right?
Constraints: Does it respect my constraints, tone, time, scope?
Test: Can I validate with a quick sanity check or counterexample?
Discernment prompts
“List your assumptions and uncertainty.”
“Attack your own answer. Find 5 ways it could fail.”
“Give me a minimal test plan to verify this in 10 minutes.”
- Description What it is: The craft of specifying intent so AI can execute cleanly. High-signal description format
Context: what we are doing and why
Objective: output and intended use
Constraints: time, tools, style, forbidden items
Inputs: data, examples, references
Definition of done: how we’ll judge success
Description template
Context:
Objective:
Audience:
Constraints:
Inputs:
Output format:
DoD (Definition of done):
Description prompts
“Ask me only the 5 questions you truly need to deliver a 9.9/10 output.”
“Convert my messy brief into a spec, then execute.”
- Diligence What it is: The boring discipline that makes results reliable. Diligence loop
Draft fast
Critique hard
Tighten scope
Verify claims
Final polish
Diligence rules
One owner, one outcome, one deadline
Keep a changelog of what changed and why
Always include a verification step for facts
Ship in iterations, not fantasies
Diligence prompts
“Turn this into a 30-minute execution plan with checkpoints.”
“Give me a revision pass focused only on clarity and structure.”
4-D Scorecard (Daily Self-Check) Rate each 1–5: Delegation: Did I delegate the right parts?
Discernment: Did I verify or challenge the output?
Description: Did I specify clearly enough?
Diligence: Did I iterate and finalize responsibly?
If any score < 3: redo the prompt with tighter description + add critique + add verification.
One-line mantra Delegate smart. Discern ruthlessly. Describe precisely. Execute diligently.
That's how you play 4-d chess with AI