What is prompt engineering?
Prompt engineering is the practice of writing and refining the instructions (the “prompt”) you give a large language model so that it produces a useful, accurate answer. The model only sees what you type, plus anything it can look up. It does not know your manager, your client, your deadline or your company’s tone unless you tell it.
For most people in India, prompt engineering is not a separate job title. It is a work skill, like writing a clear email or building a clean Excel sheet. A marketing executive in Mumbai and a finance analyst in Bengaluru both get better results from the same tool with better prompts.
If you want to understand what happens inside the model when it reads your prompt, read our explainer What is an LLM? first. It explains tokens, context windows and why models sometimes make things up.
Why do prompts still matter if AI models keep improving?
Newer models are better at guessing what you mean, but they still cannot read your mind. A vague prompt gets a generic answer. A specific prompt gets an answer you can use with light editing.
The vendors say this themselves. OpenAI’s prompt engineering guide notes that GPT models benefit from explicit, precise instructions, while reasoning models do better with high-level goals. Anthropic’s documentation lists clarity, examples, XML structure, role prompting and prompt chaining as core techniques. Google’s Workspace team built its prompting advice around persona, task, context and format.
What is a simple prompt structure for beginners?
Use five parts. You will not need all five every time, but checking against this list catches most weak prompts.
| Part | What to write | Example |
|---|---|---|
| Role | Who the AI should act as | “Act as an HR business partner at a 200-person IT services company.” |
| Task | One clear action, starting with a verb | “Write an internal announcement about our new hybrid work policy.” |
| Context | Background, audience, facts the AI cannot know | “Employees will work from the Hyderabad office on Tuesday, Wednesday and Thursday from 1 November. Many staff commute over an hour.” |
| Format | Length, structure, tone, output type | “Under 200 words, friendly but clear, with a 3-point FAQ at the end.” |
| Constraints | What to avoid, what to do if unsure | “Do not promise any travel allowance. If something is unclear, list your questions first.” |
Put together, the prompt looks like this:
Act as an HR business partner at a 200-person IT services company.
Write an internal announcement about our new hybrid work policy.
Context: From 1 November, employees will work from the Hyderabad
office on Tuesday, Wednesday and Thursday. Monday and Friday are
work-from-home days. Many staff commute over an hour.
Format: under 200 words, friendly but clear, with a 3-point FAQ
at the end.
Constraints: do not promise any travel allowance. If anything is
unclear, list your questions before writing.A quick memory aid: R-T-C-F-C. For small tasks, Task plus Format is often enough. For anything that goes to a client, a manager or a public audience, use all five.
Which prompt engineering techniques should beginners learn first?
These six techniques cover most everyday work. Each has a weak “before” prompt and a stronger “after” version.
1. Be specific about the audience and goal
Most weak prompts are short and leave the AI to guess. Tell it who will read the output and what should happen after they read it.
Before: Write a LinkedIn post about our webinar.
After: Write a LinkedIn post announcing our free 45-minute webinar
on GST e-invoicing for small manufacturers in Gujarat. Readers are
business owners, not accountants. Goal: get sign-ups. Keep it under
120 words, no hashtags except #GST, and end with a clear call to
register.2. Show an example (few-shot prompting)
When you want a specific style or format, show one or two examples. OpenAI’s guide calls this few-shot learning: the model picks up the pattern from your examples.
Before: Write product descriptions for our kurtas.
After: Write product descriptions in the same style as this example.
Example:
Name: Indigo block-print cotton kurta
Description: Hand block-printed in Jaipur on breathable cotton.
Straight cut, side slits, fits true to size. Machine wash cold.
Now write descriptions for:
1. Mustard chanderi kurta with gota border
2. White chikankari kurta, Lucknow embroidery3. Ask for a clear output format
Tables, bullet lists, fixed headings and word limits make output easier to check and paste into your work.
Before: Compare Zoho CRM and HubSpot for us.
After: Compare Zoho CRM and HubSpot for a 15-person B2B sales team
in India. Return a table with these columns: Criteria, Zoho CRM,
HubSpot, Notes. Rows: pricing model, ease of setup, WhatsApp
integration, reporting, support hours in IST. Mark anything you
are not sure about as "verify on vendor site".The last line matters. It tells the model it is allowed to be unsure, which reduces confident guesses about prices and features.
Want to turn prompts into workflows that run on their own?
Week 2 of the ISS AI & Agentic Systems program covers prompt templates, chains and a personal prompt library, and later weeks move on to automations and agents. Read the curriculum, or download the free AI Projects Starter Kit on this page to practise first.
View AI & Agentic Systems curriculum →4. Break big tasks into steps (prompt chaining)
Asking for a full report in one prompt usually gives you something shallow. Split the work into steps and check each one before moving on. Anthropic’s documentation calls this prompt chaining.
Before: Write a market research report on EV scooters in India.
After (step 1): List the 8 questions a product manager should answer
before launching an EV scooter in tier-2 Indian cities. Do not
answer them yet.
(step 2): For questions 1 to 3, tell me what data I would need and
where it is usually published.
(step 3): Here are my notes and sources [paste]. Draft section 1
of the report using only these notes.5. Separate instructions from the material
When you paste a long email thread, policy or transcript, mark clearly where it starts and ends. Both OpenAI and Anthropic recommend Markdown headings or XML-style tags for this.
Before: summarise this [pasted 3-page email thread]
After: Summarise the email thread inside the <thread> tags for a
manager who has 2 minutes. Give: the decision needed, the options,
who is waiting on whom, and the deadline.
<thread>
[paste thread here]
</thread>6. Ground the answer and allow “I don’t know”
Language models can produce confident but false answers, which OpenAI calls hallucinations. Reduce this by limiting the model to the material you provide and giving it permission to say something is missing.
Before: What is our leave policy for new joiners?
After: Answer using only the policy text below. If the answer is not
in the text, reply "Not stated in the policy" and do not guess.
Quote the sentence you used.
<policy>
[paste policy]
</policy>
Question: How many casual leaves does a new joiner get in the
first 6 months?30 copy-ready prompts for work in India
Replace the text in square brackets with your details. Each prompt already includes a role, task, format or constraint, so you can use it as is and refine from there.
Email and communication
- “Rewrite this email to a client so it is polite but firm about the overdue invoice of [amount] due on [date]. Keep it under 120 words. [paste draft]”
- “Write a leave request to my manager for [dates] for [reason]. Mention who will cover my work: [name/role]. Tone: professional, short.”
- “Turn these bullet points into a clear weekly update email for my team lead. Use three headings: Done, In progress, Blocked. [paste bullets]”
- “Reply to this customer complaint. Apologise once, explain the next step and give a timeline. Do not offer a refund. [paste complaint]”
- “Translate this message into simple Hindi that a field sales team will understand. Keep product names in English. [paste text]”
Excel and data
- “My data is in A1:D200 with headers Date, Region, Product, Sales. Write an Excel formula for total Sales where Region is North. Explain it in one line.”
- “Explain what this formula does, step by step, as if I am new to Excel: [paste formula]”
- “This formula gives #N/A for some rows. List the three most likely reasons and how to check each one: [paste formula and sample rows]”
- “Suggest 5 charts for this monthly sales data and say which question each chart answers. [paste summary]”
- “Write clear column names and a one-line description for each column in this table so a new analyst can understand it. [paste headers and 5 rows]”
Reports and documents
- “Turn these meeting notes into a one-page report with: summary, decisions, action items (owner and date), open questions. [paste notes]”
- “Create an outline for a 10-slide presentation on [topic] for [audience]. One line per slide on what it should prove.”
- “Edit this paragraph for clarity. Keep my meaning, cut it by a third, and list the changes you made. [paste text]”
- “Write a standard operating procedure for [process] as numbered steps, with a checklist at the end. Ask me questions first if a step is unclear.”
- “Read this document and list every date, amount and deadline in a table with the sentence it came from. [paste document]”
Meetings and planning
- “Draft a 30-minute meeting agenda to decide [decision]. Include time per item and what we need to agree by the end.”
- “I have these 9 tasks this week. Sort them by urgency and impact in a 2x2 table and suggest what to drop. [paste tasks]”
- “Play the role of a sceptical CFO. Ask me the 5 hardest questions about this budget proposal. [paste proposal]”
- “Create a project plan for [project] over 6 weeks as a table: week, tasks, owner, risk.”
- “Summarise this call transcript in 5 bullets, then list anything that sounds like a commitment. [paste transcript]”
Job search and career
- “Compare my resume with this job description. List the missing skills and keywords. Do not invent experience I don’t have. [paste both]”
- “Rewrite these resume bullets using the format action, task, result. Keep numbers only where I have given them. [paste bullets]”
- “Act as an interviewer for a [role] at a mid-size Indian company. Ask me one question at a time and give feedback after each answer.”
- “Write a short LinkedIn message to a [role] at [company type] asking for a 15-minute chat about their work. No flattery, under 80 words.”
- “I have offers A and B [paste details]. Build a comparison table on pay, learning, commute and growth, and list questions I should ask HR.”
Marketing and customers
- “Write 5 subject lines for a Diwali offer email to existing customers of a [business]. Under 50 characters each, no all-caps.”
- “Group these 40 customer reviews into themes, with a count per theme and one example quote each. [paste reviews]”
- “Write a WhatsApp broadcast message for [offer]. Under 60 words, one emoji at most, clear call to action.”
- “Suggest 10 blog topics for [business] that answer questions our customers ask before buying. Add the likely search query for each.”
- “Critique this landing page copy as a first-time visitor. What is unclear, what is missing, what would stop you signing up? [paste copy]”
What are the most common prompt mistakes?
- Asking for too much at once. Split big tasks into steps (technique 4).
- Leaving out the audience. “Write a summary” for a CEO and for a new intern should look very different.
- Trusting facts, numbers and links without checking. Ask for sources, open them, and compare with the original.
- Pasting confidential data into a personal account. Check your company’s AI policy first. Our guide on how to use ChatGPT for work covers the privacy settings.
- Never iterating. If the first answer misses, say exactly what was wrong (“too formal”, “missing the deadline”) rather than starting again.
How can you practise prompt engineering?
Pick three tasks you do every week, such as a status email, a report and a spreadsheet clean-up. Write a prompt for each using the five-part structure, run it, fix what is wrong and save the final version. After two weeks you will have a small, tested library for your own job.
Then go one level up. Prompts become more powerful when they are chained together and connected to tools, which is how simple automations and AI agents work. Our guides to AI automation projects for beginners and what AI agents are show that next step. If you are comparing structured options, see how to choose a prompt engineering course in India.
The ISS AI & Agentic Systems program covers prompting fundamentals and prompt chains in Week 2, with a personal prompt library as the output, before moving on to research, content, automation and agent work. It runs for 16 weeks at about 12–14 hours a week, and the fee is ₹69,999 (₹66,499 if paid upfront).
Frequently Asked Questions
What is prompt engineering in simple words?
Prompt engineering is writing clear instructions for an AI tool so it gives you a useful answer. A good prompt says who the AI should act as, what task to do, the background it needs, the format you want and anything it should avoid.
Do I need coding to learn prompt engineering?
No. Everyday prompt engineering is plain-language writing. Coding helps only when you want to call AI models from your own software or build automated workflows, and even many automation tools today are no-code.
What is the best prompt structure for beginners?
A simple five-part structure works well: role, task, context, format and constraints. For small tasks, the task and format are often enough. For anything shared with a client or manager, use all five parts.
What is few-shot prompting?
Few-shot prompting means including one or more examples of the output you want inside the prompt. The model copies the pattern, style and format from your examples, which is useful for product descriptions, email templates and data labelling.
Is prompt engineering a good career in India?
On its own, prompt engineering is usually a skill inside a wider role rather than a job title. It is most valuable when combined with domain knowledge or with skills like automation, AI agents and building AI tools.
Sources and methodology
- Google, 5 tips for writing great prompts for Gemini in the Workspace side panel: persona, task, context and format. Checked September 2026.
- OpenAI, Prompt engineering guide: Markdown and XML formatting, few-shot learning, context, GPT vs reasoning models. Checked September 2026.
- Anthropic, Prompt engineering overview: clarity, examples, XML structuring, role prompting, prompt chaining. Checked September 2026.
- OpenAI, Why language models hallucinate (5 September 2025): definition of hallucinations. Checked September 2026.
- ISS program details come from our program page and fees page.
Method: the five-part structure, the before-and-after examples and the 30 prompts were written by the ISS Editorial Team, based on the vendor guidance above. They are starting points, not tested benchmarks; results vary by tool and model.
Next steps
Start with three prompts from the list above and adapt them to your own job this week. When you are ready to build projects that use prompts inside workflows, download the free AI Projects Starter Kit on this page for six project briefs.
If you want structured, live teaching from prompting to automations and agents, read the AI & Agentic Systems curriculum. You can apply for free and pay only after you accept an offer. ISS does not guarantee jobs or placement.