📘 Using AI Effectively

Prompt Engineering + Real Practices

🔹 Introduction

AI has deeply transformed how we work, learn, and solve problems. My goal is not just to teach concepts but to guide you in changing your way of thinking. This session mixes knowledge from the Andrew Ng & Isa Fulford Prompt Engineering Course with my own daily experiences of learning and building with AI.

🧠 Two Principles of Prompting

Everything in prompt engineering is built on two powerful rules:

  • ✅ Principle 1: Write clear and specific instructions (remember: clear ≠ short)
  • ✅ Principle 2: Give the model time to think (guide it step by step)

✅ Principle 1: Clear & Specific

AI is like a smart assistant who wants to help but easily gets confused if you’re vague. Writing clearly is like giving a friend exact directions: *not just “Go over there”*, but “Walk 2 blocks north, then turn right at the café.” It’s not about making prompts short, it’s about making them unambiguous.

🛠️ Tactic 1: Use Delimiters

Delimiters are like fences that clearly mark what belongs where. They reduce confusion by separating instructions from data.

  • """ (Triple quotes)
  • ``` (Triple backticks)
  • --- (Triple dashes)
  • <> (Angle brackets)
  • <tag></tag> (XML tags)
  • Markdown formate

Example: Ask AI to summarize text inside """ so it knows exactly which part to work on.

🛠️ Tactic 2: Ask for Structure

Instead of a vague “Tell me about…” → ask AI to put its response into a format you can use directly. This saves time and avoids messy answers.

  • JSON for structured data
  • HTML for web content
  • Markdown, tables, or bullet lists for clean summaries
  • Code in specific languages when needed

🛠️ Tactic 3: Check Assumptions

Before asking the model to solve something, make sure it has the necessary context. Don’t assume it “knows” everything. If information is missing, the model may fill gaps with fantasy. Always check: “Does AI have what it needs to solve this?”

🛠️ Tactic 4: Few‑Shot Prompting

Don’t just explain the rules—show examples. When you show AI how you want the answer to look with 2‑3 samples, it usually follows that exact style in its future responses.

Example: Give it a “Q → A” pair, then ask a new question.

✅ Principle 2: Let the Model Think

AI struggles when you rush it for a final answer on a hard task. Instead, let it “talk through the problem” by prompting it to walk step by step. This is like asking a friend in math class: *“Don’t just tell me the answer, show your working.”*

⚠️ Limitation: Hallucination

Sometimes AI makes things up—confidently. It can output content that sounds plausible but is factually wrong. These are called hallucinations.

Think of it like a student bluffing in an exam: speaking smoothly, but wrong.

✅ Reducing Hallucination

  • Find accurate sources first
  • Embed relevant facts in your prompt
  • Ask AI to base answers strictly on given input

🔁 Iterative Prompt Development

Prompting is not one‑and‑done. Like scientific experiments, you try → review → refine → repeat.

  1. Start with idea
  2. Write prompt
  3. Check output
  4. Identify errors
  5. Refine and try again

📋 Common Use Cases of AI

AI shows its strength in three core use cases: Summarizing, Transforming, Expanding. Each one helps us think more clearly and work more efficiently.

🔍 Summarizing

Instead of reading long complex text, AI can condense information to key insights so you see the bigger picture quickly.

  • Inferring: detect sentiment (positive, negative, neutral)
  • Summarize long documents into short, clear notes
  • Extract action items from meeting transcripts or manuals

Example: “Summarize this report into 5 bullet points focusing on risks.”

🔄 Transforming

AI is powerful at converting knowledge from one form into another, bridging gaps between languages, formats, and complexity levels.

  • Translate languages (English ⇄ Hindi, etc.)
  • Convert code (Python ⇄ JavaScript)
  • Turn a math explanation into simple, everyday language
  • Simplify technical research into easy teaching material

Example: “Explain this Java code in a way a beginner Python learner would understand.”

➕ Expanding

AI is great at enriching short text into detailed context. Expanding is not about adding fluff—it’s about creating meaning, structure, and flow.

  • Expand short notes into full essays or articles
  • Turn rough thoughts into polished emails
  • Generate detailed reports from outlines

Example: “Take these 3 bullet points and turn them into a professional email for my team.”

💼 How I Use AI Daily

Beyond the course lessons, my personal practices and experiences show how AI truly becomes a partner in learning and work.

🧠 Transforming Ideas into Systems

Many times my head is full of vague plans. With AI, I convert raw thoughts into step-by-step systems. This turns imagination into practical action.

Example: I explain a workflow in plain text → AI helps me draft pseudocode or diagrams → I refine until it becomes working software.

📚 Learning with AI

When learning new concepts, I start from what I know and let AI fill in the gaps gradually. I don’t just ask for definitions, I discuss with AI like a mentor.

  • Ask questions until I understand deeply
  • Test my understanding by re-explaining in my own words
  • Use real-time AI voice chat for richer learning

🎙️ Voice Chat with AI

I talk to AI like a real conversation partner. It has the breadth of global knowledge, my job is to keep asking until I understand. If I don’t get it, I ask for simpler explanation until it clicks.

🛠️ Learning New Tools by Doing

I often dive straight into tools and let AI fill knowledge gaps. Instead of long tutorials, I use AI to get immediate clarity, then build as I learn.

  • Streamlit → turning Python scripts into web apps
  • SvelteKit → frontend + backend web development
  • Leveraging prior experience (React/Next.js) with AI bridging gaps

📄 Preparing Better Documents

I rely heavily on AI to draft study notes, structure documents, and make them share-ready. This applies both to personal learning notes and public docs for others.

  • Draft messy notes into clean versions
  • Restructure ideas for clarity
  • Create project documentation effortlessly

🔧 System Tasks with AI

On Ubuntu, I often need quick and efficient help for everyday tasks. AI acts as my command-line buddy.

  • Merging PDFs
  • Converting file formats (images → PDF)
  • Learning new shell commands on the go

📜 Automation with AI

I automate repetitive web tasks using tools like Selenium with help from AI. This saves huge time and effort.

  • Create data entry automation
  • Debug scripts faster
  • Arrive at efficient working code quickly

💻 Coding & Debugging

AI is my coding assistant across both familiar and new problems. Sometimes I need quick code I know but don’t want to retype; other times, I want help on bugs I’ve never seen.

  • Write known code faster
  • Debug unfamiliar errors
  • Discover better coding practices

🧾 Conclusion

Using AI is not about magic. It’s about learning how to talk with it clearly and work with it iteratively. Treat AI as a skilled partner: when you guide it step by step, it helps transform vague thoughts into real outcomes.