Module 2 - Total

• Effective prompts are the basis of prompt engineering, as AI models analyze patterns and probabilities from training data.
• Well-crafted prompts ensure AI understands expectations for desired outcomes.

How AI Interprets Input

• AI follows a rulebook rooted in data patterns.
• Ambiguous inputs lead to vague responses; clear instructions yield precise results.

Clarity

• Clarity is essential.
• Vague prompts yield generic responses, while specific prompts give targeted results.
• Example: "Write a story about a hero" is vague. "Write an inspiring 500-word story about a firefighter who overcomes their fears to save lives" is specific.

Specificity and Context

• More context leads to better results by specifying audience, format, tone, and length.
• Asking "Write a blog post about productivity" is broad. Instead, request "Write a 700-word blog post targeted at freelancers, sharing five actionable productivity tips with relatable examples."

Real-World Example

• Poor Prompt: "Describe marketing strategies."
• Improved Prompt: "Explain three effective marketing strategies for tech startups, including how to leverage social media and email campaigns with specific examples."
• Always add specific parameters like tone, word count, format, or audience to guide AI to the desired outcomes.

Direct Prompting

• Direct prompting is a no-nonsense approach ideal for structured tasks requiring clear, concise outputs.
• Example Prompt: "List five key benefits of remote work for millennials in bullet points."
• Use for creating step-by-step guides, generating specific lists and summarising reports.

Conversational Prompting

• Conversational prompts mimic a dialogue tone, suitable for creative or brainstorming tasks.
• Example Prompt: "I’m planning a novel set on Mars. What would be a compelling motivation for the protagonist to leave their colony?"
• Use for brainstorming creative ideas, storytelling, and exploring open-ended possibilities.

Practical Tip

• Experiment with both styles of prompting.
• Notice the tones and depth of ideas generated by comparing prompts like "Create a list of 10 blog ideas for health coaching" and "What are some trending topics a health coach could blog about to attract millennials?"

Analysing Outputs

• Review the AI's initial response critically, ensuring instructions were specific enough.

Using Follow-Up Prompts

• AI can function like a teammate to clarify instructions.
• Example: "Can you expand on point three with more examples?"

Avoiding Common Mistakes

• Avoid vague, multi-tasked, or contradictory prompts.
• Break prompts into smaller tasks to improve responses.

Pro Tip

• Document prompts that consistently deliver strong outputs, creating a “prompt library” for repeated use.

Interactive exercise

• Start with a broad prompt, like "Tell me about AI".
• Refine it incrementally until it provides a specific, actionable response, for example, "Explain how AI is transforming supply chain management, with examples of current applications."

Tone Control

• Use tone instructions like “Write in a professional tone suitable for a white paper” or “Write in a friendly tone for a blog aimed at millennials.”

Add Constraints

• Provide word counts or format specifications.
• Example: “Write a 200-word summary in bullet points with three subheadings.”

Custom Outputs

• Ask for unique formats like tables, outlines, or summaries.
• Example Prompt: "Draft a Grade 7 lesson plan on renewable energy, including objectives, materials, and hands-on activities."

Advanced Applications

• Marketing: Create engaging headlines for ads and social media.
• Education: Design interactive and age-appropriate lesson plans.
• Business Planning: Generate executive summaries targeting C-suite audiences.

Vagueness

• Less detail leads to weaker results.
• Solution: Add details such as audience, tone, format, and examples.

Overloaded Prompts

• Attempting multiple tasks in one prompt often causes confusion.
• Solution: Split complex tasks into smaller steps.

Ignoring AI Limitations

• Generative models don’t “understand” content; they predict it.
• Solution: Be prepared to fact-check and guide with follow-up instructions.

Actionable Tip

• Test prompts across multiple platforms (e.g., GPT, Claude, Jasper) to learn their strengths.
• Compare how each platform handles a task, such as “Draft a mission statement for a startup focusing on sustainable energy.”