Last week I introduced Context Debt — the invisible tax you pay every time AI guesses instead of knows.
This week: proof that it's real. Same AI. Same request. Two outputs you'd never believe came from the same model.
Here's the prompt both versions got:
"Write me a client proposal for a consulting engagement."Version A — the bare prompt.
ChatGPT returned three paragraphs about "leveraging strategic synergies to drive transformational outcomes." It mentioned "stakeholders" four times. It suggested a "phased approach" with no phases defined. The pricing section said "competitive rates tailored to your needs."
You'd never send it. You know it. I know it. The model doesn't know — because nobody told it what "good" looks like.
Version B — same prompt, with a Context Stack pasted first.
The Context Stack told the model six things:
Role: Senior management consultant, 12 years in operational efficiency for mid-market SaaS companies.
Objective: Win a 6-month engagement with a 200-person SaaS company struggling with sales-marketing alignment. Target contract: $120K.
Business context: Client's marketing team generates leads that sales ignores. Sales blames lead quality. Marketing blames follow-up speed. CEO is frustrated.
Constraints: No travel required. Must integrate with their existing HubSpot instance. Board wants measurable ROI within 90 days.
Output format: Executive summary, diagnostic framework, phased timeline, deliverables per phase, pricing with payment terms, risk mitigation.
Guardrails: No jargon. No "synergy." No vague promises. Every deliverable must be tied to a metric.
Same one-line prompt. The output: a structured proposal with a specific diagnostic framework, three clearly defined phases, deliverables mapped to quarterly KPIs, $120K pricing with milestone-based payments, and a risk section that addressed the CEO's board pressure directly.
Minor edits. Ship it.
That's the difference. Not a better prompt. Not a smarter model. Not a $200/month subscription to a fancier AI tool.
A 25-minute document that tells the AI who you are, what you're trying to accomplish, and what the constraints are. Build it once. Use it on every serious interaction. The output shifts from "rewrite everything" to "tweak and send."
The prompt was 20% of the equation. The context was the other 80%.
Try this before your next deadline:
Take your most common task — the proposal you write monthly, the brief you draft weekly, the outreach you send daily. Before you prompt, write down:
Who you are and what you specialize in (one sentence)
What the specific outcome needs to be (not "write something good" — the actual target)
One constraint the AI doesn't know about
Paste those three lines before your prompt. Run it. Compare the output to what you normally get.
That's three of the six layers. And it takes under five minutes.
The complete framework I use with enterprise clients — adapted for solo operators and small teams. Build it in 60 minutes. Use it forever.
Not a prompt pack. Not a course. The actual system.
See you next week.
— Chris
