---
title: "Context Changes Everything"
description: "The same question asked properly — attachments, constraints, and what 'good' looks like."
order: 2
duration: "10 min"
chapter: "01-the-org-age-of-ai"
type: lesson
---

## The same question, done properly

Before we name the problem, see how much of it you can fix right now, with nothing but the attach button. Ask the same question again — but this time, give Claude what a competent new hire would get on their first day.

Sarah tries again. This time she attaches an export of her chart of accounts (a CSV her accountant sent her once) and writes:

> "Here's our chart of accounts. A contractor did $4,800 of work — replacing the walk-in cooler compressor at the Main Street store and repainting the Oak Avenue storefront. How should I split and categorize this payment? Also, anything I should watch out for?"

The answer transforms. Claude now routes the compressor to 6250 — Building Improvements, because it can see that account exists and what it is for. It splits by location because Sarah mentioned two stores. It even asks a genuinely useful question back: "Is this contractor incorporated? If not, and you've paid them over $600 this year, this payment has 1099 implications."

Same model. Same day. The difference is entirely in what Sarah gave it. This is the first craft skill of working with AI, and most people never learn it: **the quality of the answer is bounded by the quality of the context.** Attach the document. Paste the numbers. Name the constraints. Say what "good" looks like.

But notice what is still missing. Claude only knew about the two locations because Sarah typed it. It only saw the chart of accounts because she attached it. Tomorrow, in a new conversation, it will know none of this — she would have to attach and explain everything all over again. And the 1099 running total for that contractor? That lives in QuickBooks, and no attachment fixed it.

<div class="exercise">
  <div class="callout-label">Try This</div>
  <p>Re-ask your question from the previous lesson, but this time do it properly: attach the relevant document (a report, a policy, a spreadsheet — whatever a new hire would need), state the constraints, and say what a good answer looks like. Compare the two answers. Then write down what was <em>still</em> missing — the things you couldn't fix with an attachment because they live in a system, or in someone's head, or in accumulated history.</p>
</div>

The things you could fix with context are a prompting problem, and you just solved it. The things you could not fix are an *organizational* problem. That gap — between what you can paste into a chat and what your organization actually knows — is what this entire course is about.

## Check your understanding

Answer in your own words — write it down before opening the key. Your tutor grades against the criteria and generates fresh variants on retries.

**Q1.** A colleague asks Claude "how should I price my consulting services?", gets a generic answer, and concludes AI is overhyped. What would you tell them to do differently on the next attempt — and what will *still* be missing from the answer even if they do it perfectly?

<details>
<summary>Answer key — attempt every question first</summary>

## Answer key

### Q1

**Model answer:** Give Claude what a competent new hire would get: attach the rate sheet or past proposals, state the constraints (market, positioning, target clients), and say what a good answer looks like. Even done perfectly, the answer will still miss what lives outside the conversation: their actual historical win/loss pricing, the unwritten rules about which clients get discounts, and anything stored in systems Claude cannot see.

**Pass criteria:** names attaching real documents/context AND stating constraints; identifies that org-specific, undocumented, or system-resident knowledge is still missing

</details>


**Next:** [The Gap No Prompt Can Close](/mastering-claude/01-the-org-age-of-ai/03-the-gap/)