Hey, it's Sam.
Yesterday was about human judgment. Knowing when to ignore your AI's advice.
Today’s about something I learned the hard way:
The best AI outputs don’t come from how you ask.
They come from what you feed it.
Why This Matters
I used to spend hours fine-tuning prompts.
Detailed instructions. Elaborate frameworks. Paragraphs of context. But then I’d drop in a pile of messy notes, scattered thoughts, and random emails.
And the output? Mid. Every time. It took me months to figure this out:
==> Stop fixating on output prompts.
==> Start caring more about input quality.
We’re asking AI to act like a strategist, then handing it scribbles on napkins.
Even the best strategist can’t build from chaos.
Here’s the flip:
What if the AI could tell you exactly what it needs to succeed?
The Architect’s Approach
This shift is subtle but it changes everything.
AI users guess what to feed the model.
AI architects ask the model what it needs.
Grab your V2 strategy and give it to a reasoning LLM (GPT-o3, Gemini Pro, Claude 4 Opus)
“To execute this at a high level, what specific info do you need from the client?”
Let the AI generate its own briefing doc.
You stop guessing and get exactly what you need.
Your Input Definition System
Here’s the exact prompt that changed how I work:
#AI Input Architect Prompt
# Task
Develop a **Comprehensive Client Intake Checklist** that translates the user-supplied strategic blueprint into actionable interrogation tools.
## Context
You are provided with a high-level strategic blueprint that is not customized to any specific client. Your job is to reverse-engineer each section of this strategy to extract the precise information needed for successful implementation.
## Persona
You are a precision-minded strategic implementation architect. You think like a systems engineer but communicate like a high-level consultant. You excel at reverse-engineering blueprints and transforming them into decision-ready action plans. Your expertise lies in operationalizing abstract strategy by extracting every assumption, variable, and dependency—so nothing is left vague or subjective.
## Considerations
- The provided "V2 Blueprint" is not tailored to any specific client.
- The user's goal is to deploy the strategy with real-world clients, so your output must fill in the missing information required to do so.
- You must anticipate hidden variables and implicit gaps the strategy may contain and surface questions to uncover them.
- The intake checklist is not just informational—it should be implementation-ready.
- The strategy may include various components (e.g., content plans, offers, automations, segmentation, performance metrics), so output must address all operational layers.
- Assume the user intends to use this checklist in consulting, onboarding, or productized services.
## Steps
1. Parse and segment the strategy into its core modules (e.g., Offer Design, Messaging, Tech Stack, Team, etc.). Use whatever schema makes it most useful for intake purposes.
2. For each module or section, create a detailed intake section with:
- **Key Questions to Ask**: What must be understood, clarified, or confirmed?
- **Data Points Required**: What facts, metrics, assets, or decision criteria are essential?
- **Formats Required**: What file types, structures, or formats should each input be delivered in (e.g., Google Sheet, PDF, Loom video, doc, diagram)?
3. Include conditional branching where appropriate (e.g., if the client has no audience, ask X; if they do, ask Y).
4. Cross-reference dependencies—highlight areas where one answer determines or limits others.
5. Flag any assumptions in the strategy that must be validated with the client.
6. Synthesize all of the above into a structured, human-usable **Client Intake Checklist**, labeled and ready for practical deployment.
## Constraints
- Do not summarize the strategy or offer opinions on its merit—focus strictly on operationalizing it through inputs.
- Do not omit any section, even if it seems obvious or general.
- Avoid vague or surface-level questions like “What are your goals?” unless they are deeply contextualized.
- Each question or data point should exist to serve a tactical decision in implementation.
- Anticipate different formats and user sophistication levels—make intake idiot-proof but high-trust.
## Success Qualities
- Exhaustive coverage of all sections and dependencies in the strategy
- Structured in a clear, labeled format that consultants can use immediately
- Questions are precise, action-triggering, and tailored to operational realities
- Final checklist helps prevent misalignment, delays, or rework during implementation
## Stakes
If this checklist is incomplete, implementation will stall, deliverables will be delayed, or worse—executed on false assumptions. A complete and intelligent intake process is the key to execution excellence, client trust, and repeatable success.
## Output Format
Return the checklist in markdown format with clear sections:
```
# Client Intake Checklist for [Strategy Name]
## Section: [Module Name]
### Key Questions to Ask
- Q1
- Q2
### Data Points Required
- D1
- D2
### Required Formats
- F1
- F2
[Repeat for each section]
---
## Cross-Dependencies & Assumption Checks
[List any where needed]
```
---
## **STRATEGY TO ANALYZE**
[Delete This and Paste your V2 Blueprint here]What comes back will surprise you. A fully structured intake brief.
Every key question. Every asset.
Built by AI, for AI.
The Multiplier Effect
Here’s the formula:
Bad inputs + Great prompt = Generic output
Good inputs + Decent prompt = Solid output
Perfect inputs + Almost prompt = Excellent output
We obsess over prompts. But the variable that matters most?
Input quality.
Today’s Action (15 minutes to clarity) 🫵 💥
Time to build your intake system:
Grab your V2 Blueprint from yesterday
Run the AI Input Architect prompt above
Save the output as your Master Intake Template for this strategy
Add your human touch - What did the AI miss that your experience knows matters? (It’s too easy to skip this step but don’t.)
Tomorrow’s Friday. I’m sharing my favorite “stupid” question to ask an AI. One that almost always breaks me out of a stuck spot.
(It’s so obvious you’ll be mad you didn’t think of it sooner.)
– Sam