Module 01
What is Prompting?
A prompt is your instruction to an AI. The quality of your output is directly determined by the quality of your input. This is the most learnable skill in AI.
Why prompting matters
Most people use AI the same way they use Google — type a few words, read the first result. But AI language models are not search engines. They are extraordinarily capable systems that respond to the precision and richness of your instruction.
| Weak prompt | Output |
|---|---|
| "tell me about SIP" | Generic Wikipedia-style explanation |
| "write a cover letter" | Bland template with [YOUR NAME] placeholders |
| "analyse this company" | Surface-level overview |
| Strong prompt | Output |
|---|---|
| "Explain SIP investing to a 25-year-old earning ₹60,000/month who has never invested before. Cover what it is, how to start, how much to invest, and the top 3 mistakes beginners make. Use simple language and Indian rupee examples." | Specific, actionable, India-relevant guide |
The prompting mindset
Think of AI as an extraordinarily intelligent assistant who has just joined your team. They are brilliant — but they don't know your context, your audience, your style, or your constraints unless you tell them.
- The more context you give, the better the output
- Iterate — good prompting is a conversation, not a one-shot request
- Be specific about format: "give me a numbered list of 5 points, each under 2 sentences"
- Tell it who to write for: "explain this to a 50-year-old with no finance background"
What AI models are actually doing
When you send a prompt, the model predicts the most likely next token given your input. It is not "thinking" — it is pattern-matching at enormous scale. Understanding this explains both its power and its limitations:
- Power: It has seen millions of examples of great writing, clear explanations, and correct code — so it can produce all of these
- Limit: It will confidently produce plausible-sounding but wrong answers if your prompt steers it toward unfamiliar territory — always verify
Module 02
Anatomy of a Great Prompt
Six ingredients that transform an average prompt into an exceptional one. You don't need all six every time — but knowing each one changes what you can get.
The six prompt ingredients
| # | Ingredient | What it does | Example |
|---|---|---|---|
| 1 | Role | Sets the persona and expertise level | "Act as a senior investment banker with 15 years of experience" |
| 2 | Context | Gives background so the AI understands your situation | "I am a 28-year-old software engineer earning ₹18L/year with no debt" |
| 3 | Task | The specific thing you want done | "Create a 3-year investment plan that allocates my monthly surplus" |
| 4 | Format | How the output should be structured | "Present as a table with columns: asset class, allocation %, monthly amount, rationale" |
| 5 | Constraints | Limits and requirements | "Assume 12% equity return, 7% debt return. No crypto. India context." |
| 6 | Audience | Who will read/use the output | "Write for someone who has never invested before — avoid jargon" |
Before and after: same request, different results
Weak
"write a linkedin post about financial planning"
Strong
"You are a personal finance educator writing for Indian professionals aged 25–35. Write a LinkedIn post about the importance of starting SIP investing in your 20s. Tone: conversational but authoritative. Format: hook (1 line) + 3 key insights as short paragraphs + CTA. Max 250 words. Avoid clichés like 'financial freedom'. Include one real number example."
The role prompt: your most powerful lever
Assigning a role dramatically shifts the quality, depth, and style of AI output. Use roles that match the expertise you need:
| Role | When to use |
|---|---|
| "Act as a CFO reviewing this business plan" | Business/finance analysis — get financial rigour and risk flags |
| "Act as a senior recruiter at Goldman Sachs" | CV/cover letter review — get hiring manager perspective |
| "Act as a hostile investor trying to find holes in this thesis" | Investment research — adversarial stress-testing |
| "Act as a teacher explaining to a 10-year-old" | Any complex concept — guaranteed plain English |
| "Act as a copywriter who specialises in high-conversion emails" | Marketing copy — gets specific, persuasive output |
The format instruction: always specify structure
Without a format instruction, AI defaults to paragraphs. Specifying format dramatically improves usability:
- "Give me a table with columns: X, Y, Z"
- "Number each point. Max 3 sentences per point."
- "Use headers for each section. Keep the total under 400 words."
- "Give me the answer in JSON format with keys: title, summary, pros, cons"
- "Format as a step-by-step checklist I can print and use"
Module 03
Core Prompting Techniques
Five proven techniques used by AI engineers and power users. Each one solves a different problem.
1. Zero-shot prompting
Ask the AI to do something with no examples — just a clear instruction. Works for most common tasks.
Example
"Summarise the key risks in this paragraph in 3 bullet points: [paste text]"
Use when: the task is clear and well-defined. Fails when: the task is ambiguous or requires a specific style the AI hasn't guessed correctly.
2. Few-shot prompting
Give 2–3 examples of the input-output pair you want before asking for the real thing. The AI learns the pattern from your examples.
Example
"Rewrite these sentences to be more concise and active voice:
Original: 'The report was reviewed by the team and issues were identified.'
Rewrite: 'The team reviewed the report and identified issues.'
Original: 'It is recommended by experts that investors should diversify.'
Rewrite: 'Experts recommend investors diversify.'
Now rewrite: 'The decision was taken by the committee that all bonuses would be deferred.'"
Use when: you have a specific style, format, or transformation in mind that's hard to describe in words.
3. Chain-of-thought prompting
Ask the AI to reason step-by-step before giving its final answer. This dramatically improves accuracy on complex reasoning tasks — maths, analysis, multi-step decisions.
Without CoT
"Should I invest in index funds or active funds?"
With CoT
"Should I invest in index funds or active funds? Think through this step by step: first consider the evidence on active vs passive returns in India, then consider expense ratio impact over 20 years, then consider my situation (I have ₹10,000/month to invest, 20-year horizon, no time to research individual funds). After working through each factor, give your recommendation."
4. Persona + audience framing
Combine a role (who the AI is) with an audience definition (who the output is for). This is the fastest way to get tone-perfect content.
Example
"You are a chartered accountant explaining the new Indian income tax regime. Your audience is a 30-year-old salaried employee who has never done their own taxes and finds financial documents confusing. Explain the difference between old and new regime in 5 bullet points. Use simple language, include a worked example with a ₹12L salary."
5. Iterative refinement
The single biggest difference between average and expert AI users: they iterate. Treat your first prompt as a draft, then refine.
| Round | What you do |
|---|---|
| Round 1 | Send your best prompt — get a response |
| Round 2 | "Make the tone more conversational and cut the length by 30%" |
| Round 3 | "Add a concrete example for the second point. Change the opening line." |
| Round 4 | "Good. Now rewrite section 2 to be more specific about India." |
Pro tip: At any point you can say "What additional information would help you give me a better answer?" — the AI will tell you exactly what context it's missing.
Module 04
Prompt Library by Category
Copy, customise, and use. Real prompts that work — across finance, career, content, research, and everyday tasks.
Finance prompts
Analyse my financial situation
"Act as a certified financial planner. I am [age], earning ₹[X]/month take-home. My monthly expenses are ₹[Y]. I have ₹[Z] in savings, [list any debts]. I want to [goal: buy a house / retire early / build wealth / save for child's education]. Analyse my current financial position and give me a prioritised 3-step action plan with specific amounts."
Understand a mutual fund
"Explain the [fund name] mutual fund in plain English. Cover: what it invests in, who it is best for, the main risks, expense ratio impact over 10 years on a ₹5,000/month SIP, and how it compares to a Nifty 50 index fund. Use a table for the comparison."
Decode a financial document
"[Paste the document text or key section]. Explain this in plain English. What are the 3 most important things I need to understand? What action (if any) do I need to take? Flag any red flags or hidden fees."
Build a savings plan
"I want to save ₹[amount] in [X] years for [goal]. I can invest ₹[Y]/month. Assume [Z]% annual return. Show me: (1) whether ₹Y/month is enough, (2) what return I need if it isn't, (3) a month-by-month savings tracker table for the first 12 months."
Career prompts
Tailor your CV to a job description
"Here is my current CV: [paste CV]. Here is the job description: [paste JD]. Identify the 5 biggest gaps between my CV and what they are looking for. Then rewrite my experience bullet points to better match the JD's language and priorities. Keep everything truthful — just rephrase and reorder."
Prepare for an interview
"You are a senior hiring manager at [company] interviewing for a [role] position. Ask me the 10 most likely interview questions for this role, one at a time. After each of my answers, give me feedback: what was strong, what was weak, and how to improve it. Start with Q1."
Negotiate a salary
"I have received an offer of ₹[X] from [company type] for a [role] in [city]. My current CTC is ₹[Y] and I have [Z years] of experience. I want to negotiate to ₹[target]. Write me: (1) the exact script for the negotiation call, (2) how to respond if they say no, (3) counter-offer options if base salary is fixed."
Write a LinkedIn post that gets engagement
"Write a LinkedIn post about [topic/lesson/experience]. Tone: personal, honest, not salesy. Structure: start with a counterintuitive hook, share the story or insight in 3–4 short paragraphs, end with a question to drive comments. Target audience: [describe]. Max 200 words. No hashtags in the body."
Content creation prompts
Write a YouTube script
"Write a 60-second YouTube Short script about [topic]. Format: Hook (5 sec, startling fact or question) → Main content (40 sec, 3 key points) → CTA (15 sec, tell them to follow for more). Conversational tone. Written for speech, not reading — short sentences, natural pauses marked with [pause]. Target: Indian audience aged 22–35."
Create a social media content calendar
"Create a 2-week social media content calendar for [brand/topic] targeting [audience]. Platforms: [Twitter/LinkedIn/Instagram]. For each day include: platform, post type (tip/story/question/infographic idea), topic, and a draft caption under 150 characters. Format as a table."
Generate a carousel idea
"Design a 7-slide LinkedIn carousel about [topic]. For each slide give: slide number, headline (max 8 words), 2–3 body bullet points, and visual suggestion. Slide 1 must be a hook that makes people swipe. Slide 7 must be a CTA. Target: [audience]."
Research & analysis prompts
Stress-test an idea or decision
"I am considering [decision/investment/business idea]. Play devil's advocate. Give me the 5 strongest arguments AGAINST this decision, the 3 scenarios under which this goes badly wrong, and the early warning signs I should watch for. Be harsh — don't validate me."
Summarise a long document
"[Paste document]. Summarise this in 3 tiers: (1) The one-sentence summary, (2) The 5 most important points as bullet points, (3) Any action items or decisions required from me. Flag anything surprising, risky, or that contradicts common understanding."
Compare options objectively
"Compare [Option A] vs [Option B] for [my situation: describe context]. Create a table with the following criteria: [list 5–6 criteria relevant to your decision]. Score each option 1–5 on each criterion. Then give a weighted recommendation based on my priorities: [list what matters most to you]."
Everyday productivity prompts
Write a difficult email
"Write a professional email to [recipient: boss/client/colleague] about [situation: asking for raise/declining request/following up/pushing back on a deadline]. Tone: [direct but polite / firm / apologetic]. Key points to include: [list]. Keep it under 150 words. No filler phrases like 'hope this finds you well'."
Learn anything fast
"I want to understand [topic] in 30 minutes. I already know [what you know]. Teach me using the Feynman technique — explain it as if I am a smart 16-year-old. Start with the core concept, give one strong analogy, show how it connects to something I already know ([relate to familiar topic]), then give me 3 questions to test my understanding."
Turn a wall of text into a decision
"[Paste text]. My goal is [what you are trying to decide or do]. Extract only the information relevant to my goal. Ignore everything else. Format: 3 bullet points of key facts + my recommended next action."
Module 05
Advanced Prompting
Techniques used by professionals who get consistently exceptional results — system prompts, context management, and avoiding the most common failure modes.
System prompts: your permanent instruction layer
Most AI interfaces (Claude, ChatGPT) allow you to set a system prompt — a persistent instruction that applies to every message in the conversation. This is how you get consistent tone, style, and behaviour without repeating yourself.
Example system prompt
"You are a financial education assistant for Moneykar.com. Your audience is Indian professionals aged 25–40. Always: use Indian context (₹, Indian regulations, Indian products like PPF/NPS/ELSS), keep language simple and jargon-free, use tables and bullet points rather than long paragraphs, include a concrete numerical example wherever possible. Never recommend specific funds or stocks by name."
Context management in long conversations
AI models have a finite context window — in long conversations, early context can be "forgotten." Techniques to manage this:
- Summarise periodically: "Summarise what we've established so far in 5 bullet points" — then reference that summary in new messages
- Re-anchor: Start important new sections with "Remember: [key constraint or context from earlier]"
- Start fresh conversations: For new tasks, a fresh context is cleaner than a polluted long conversation
- Use structured handoffs: "Here is what we've decided: [bullet list]. Now move on to [next phase]"
Avoiding hallucinations in high-stakes tasks
| Technique | How it works |
|---|---|
| "If you are not certain, say so" | Instructs the model to flag uncertainty rather than fabricate confidently |
| "Only use information from the document I pasted — do not use outside knowledge" | Constrains the model to a specific source |
| "After answering, list any assumptions you made" | Surfaces hidden assumptions for you to verify |
| "Give me 3 possible answers and your confidence level for each" | Forces the model to show its uncertainty range |
| Cross-check with Perplexity (cited sources) | Verify factual claims against real sources externally |
The meta-prompt: ask AI to write your prompt
One of the most underused techniques: ask AI to help you write a better prompt for the task you want to accomplish.
Meta-prompt example
"I want to use AI to [describe your goal]. Write me the ideal prompt to accomplish this, using best practices for role, context, format, and constraints. Then explain why you wrote it the way you did."
The compound effect: Every prompt you write makes you better at the next one. Within 30 days of deliberate practice — trying different techniques, noting what works — you will be in the top 5% of AI users. Most people never move beyond zero-shot prompting.