Teacher’s Edition
Module 06
Thinking Out Loud
When and how to activate chain-of-thought reasoning. For simple tasks, it adds latency. For complex decisions, it catches errors the fast version misses. A hiring plan that falls four features short of target. A pricing strategy with a hidden margin gap.
Charter Oak Strategic Partners · Claude Mastery Program · Version 1.0 · Confidential · Not for distribution to participants
When you tell Claude to “think step by step,” you are activating chain-of-thought reasoning. Claude breaks the problem into stages, works through each one, and reaches a conclusion grounded in the intermediate steps. For simple tasks (rewrite an email, translate a phrase, name a capital), this adds latency without improving the output. For complex tasks (multi-variable decisions, financial modeling, strategic trade-offs), it is the difference between a shallow answer and one that catches errors, surfaces hidden assumptions, and reaches conclusions the quick version misses.
Extended thinking is the more advanced version. Claude allocates dedicated reasoning time before generating the response. The user does not see the thinking process in the output, but the quality of the answer reflects it. For the purposes of this module, “chain-of-thought” and “think step by step” are the accessible entry points. Extended thinking is available in Claude’s API and higher-tier plans. Mention it as a concept; do not spend time configuring it.
Opening — 5 minutes
“When you ask Claude a simple question, you want the answer. When you ask Claude a complex question, you want the reasoning. This module teaches you when to ask for each.”
“The technique is simple. Add ‘think step by step’ or ‘walk me through your reasoning’ to a complex prompt. Claude will break the problem into pieces, work through each one, and show you how it arrived at the answer. For complex problems, this catches mistakes that the quick version misses.”
“We are going to see three problems. Each one, I will show without chain-of-thought first, then with. The difference will be obvious.”
Live Demo: Three Problems — 25 minutes
demo-data/module-06/chain-of-thought-problems.md— Three problems with prompts, expected outputs, and counter-examples.Problem 1: Vendor Decision (Simple). Two vendors, different price points and delivery speeds. The quick version picks a vendor based on surface comparison. The step-by-step version calculates total costs including delivery speed premiums, quantifies the opportunity cost of slower delivery, and produces a recommendation grounded in numbers. Difficulty: low. Purpose: warm-up, shows the pattern.
Problem 2: Hiring Plan (Medium). A company needs to ship 20 features by end of year. Current capacity: 4 features/quarter. They plan to hire 4 engineers in Q1. The quick version says “4 hires should work.” The step-by-step version models the quarterly ramp (new hires take a quarter to reach full productivity), discovers the math falls short (the team can only deliver 16 features, not 20), and flags the 4-feature gap before the company commits $660K in salaries. This is the demo that makes the room sit up. The quick version of Claude gives bad advice that a CFO might act on. The step-by-step version catches the error.
Problem 3: Pricing Strategy (Complex). A competitor cut prices 15%. Should we match? The quick version says match the cut. The step-by-step version calculates the volume increase needed to maintain margin at the lower price (typically enormous), identifies the cannibalization risk to premium product lines, evaluates whether the competitor is attacking the same segment or a different one, and proposes three alternative strategies beyond a price match. This problem demonstrates that chain-of-thought does not just compute better. It thinks wider. It considers second-order effects the quick version ignores.
After running Problem 2 with and without chain-of-thought:
“Look at the quick version. It said four hires should be enough. If your CFO saw that and approved the hires, what would happen?”
Pause. Let someone answer.
“You would reach Q4 and be four features behind schedule. $660,000 spent. Target missed. And the quick version told you everything was fine.”
“Now look at the step-by-step version. It modeled the ramp. New hires do not produce at full speed in their first quarter. Claude caught that. It recalculated and flagged the gap. That is the difference between a quick answer and a careful one.”
“What did the step-by-step version catch that the quick version missed?”
For Problem 1: the total cost calculation including opportunity cost. For Problem 2: the productivity ramp that makes 4 hires insufficient. For Problem 3: the cannibalization risk and the volume math that makes matching the price cut unprofitable.
Counter-Demo: When NOT to Use It — 5 minutes
“Chain-of-thought is not a default. It is a tool. Here are three tasks where it adds nothing.”
Run the three simple tasks from the bottom of the demo file: rewrite an email, translate a sentence, name a capital city. Show that the output is the same with or without chain-of-thought, but the step-by-step version takes longer.
“When the task has one right answer and no variables, skip the step-by-step. When the task has multiple variables, requires calculation, or involves trade-offs, use it. That is the rule.”
Paired Exercise — 15 minutes
“Pairs. Think of a complex decision you face at work. Something with multiple variables. Write it as a prompt. Send it without ‘think step by step.’ Then send it again with that instruction. Compare the outputs.”
“Your deliverable: one sentence. ‘Here is what the step-by-step version caught that the quick version missed.’ If the answer is ‘nothing,’ your problem was not complex enough. Pick a harder one.”
Debrief — 10 minutes
“Who found something the step-by-step version caught that the quick version missed?”
Take three or four examples. The pattern will be consistent: the step-by-step version surfaces assumptions, catches math errors, identifies second-order effects, or flags risks that the quick version glosses over.
Then: “When would you use this at work? Name one real decision where you want Claude to show its work.”
| Segment | Activity | Time |
|---|---|---|
| Opening | Concept framing | 5 min |
| Demo | Three problems, with and without CoT | 25 min |
| Counter-Demo | When NOT to use chain-of-thought | 5 min |
| Exercise | Paired real-problem comparison | 15 min |
| Debrief | Discussion | 10 min |