Module 3 — Advanced Prompting Strategies

Chain-of-thought (CoT) prompting instructs the model to show its reasoning step by step before reaching a conclusion. This dramatically improves performance on tasks requiring logic, math, or multi-step analysis.

Simple prompt: "Is 17 a prime number?" CoT prompt: "Is 17 a prime number? Think step by step."

The CoT version produces: "To determine if 17 is prime, I need to check if it's divisible by any number other than 1 and itself. 17 ÷ 2 = 8.5 (not divisible), 17 ÷ 3 = 5.67 (not divisible), 17 ÷ 4 = 4.25 (not divisible). I only need to check up to √17 ≈ 4.1. Since 17 isn't divisible by 2, 3, or 4, it is prime."

Why does CoT work? By forcing the model to generate intermediate reasoning steps, it creates a more structured path through the problem space, reducing the chance of reasoning errors.

Tree-of-thought extends CoT by exploring multiple reasoning paths simultaneously, evaluating each, and selecting the most promising. Self-consistency generates multiple CoT solutions and takes the majority answer, improving reliability.

Structured output formatting is another advanced technique: requesting responses in JSON, markdown tables, or specific formats. This is essential for building AI-powered workflows where outputs need to be parsed by other systems.

Prompt chaining breaks complex tasks into sequential steps, where the output of one prompt becomes the input to the next. This is more reliable than asking the model to handle everything in a single prompt.

Example workflow for analysing a business report:

Step 1 (Extract): "Read this report and extract: revenue figures, key metrics, and notable changes. Output as a structured list."

Step 2 (Analyse): "Given these extracted metrics [output from Step 1], identify the three most significant trends and explain what's driving each."

Step 3 (Recommend): "Based on these trends [output from Step 2], generate three strategic recommendations for the executive team, including expected impact and implementation timeline."

Each step is focused, manageable, and produces high-quality output that feeds into the next. This is far more reliable than a single prompt like "Analyse this report and give me recommendations."

Prompt chaining enables building sophisticated AI workflows — multi-step processes where AI handles each stage. Examples: customer support ticket triage → categorisation → response drafting → quality check; or: research question → literature search → synthesis → report generation.

The key principle: decompose complex tasks into simple, well-defined steps. Each step should have a single clear objective, specific input, and defined output format.