Building Your Own AI Workflow

You have learned how LLMs work, how to write effective prompts, how to use structured patterns, and how to apply AI to research, writing, coding, and study. You understand the risks — hallucinations, bias, over-reliance — and the ethical considerations around transparency and integrity.

Now it is time to bring it all together. This final lesson is about building a personal AI workflow — a set of habits, templates, and decision-making frameworks that let you use AI consistently, effectively, and responsibly in your daily work and learning.

---

Step 1: Choose the Right Tool for the Task

Not all AI tools are equal, and not every task benefits from AI.

When AI Adds Value

• Tasks with a "blank page" problem: Brainstorming, outlining, getting started on a draft
• Mechanical text work: Grammar checking, reformatting, translating, summarising
• Learning and explanation: Understanding new concepts, getting personalised explanations
• Pattern-based generation: Boilerplate code, standard email formats, data transformation
• First-pass review: Catching obvious errors in code, text, or data before human review

When AI Adds Less Value (or Risk)

• Tasks requiring verified accuracy: Medical, legal, or financial decisions
• Highly creative or personal work: Where your unique voice and perspective are the point
• When learning is the goal: If the assignment is designed to develop your skills, AI shortcuts undermine the purpose
• Confidential or sensitive material: Where data privacy is a concern

Tool Selection

Different AI tools have different strengths:

Tool Type	Best For
General chatbots (ChatGPT, Claude)	Writing, research, brainstorming, explanation, coding
Code-specific tools (GitHub Copilot, Cursor)	In-editor code generation, autocomplete, refactoring
Image generators (DALL-E, Midjourney)	Visual content, concept art, design mockups
Specialised tools (Grammarly, Otter.ai)	Specific tasks like grammar or transcription, often with higher accuracy

Choose the right tool for the task rather than using one AI for everything.

---

Step 2: Create Reusable Prompt Templates

If you find yourself doing the same type of task repeatedly, create prompt templates that you can reuse and refine over time.

Example Templates

Meeting Summary Template:

"Summarise the following meeting notes into: 1. Key decisions made (bullet points) 2. Action items with owners and deadlines 3. Open questions that need follow-up

Keep it concise — no more than one page.

Meeting notes: [paste notes]"

Code Review Template:

"Review the following code for: 1. Bugs or potential errors 2. Security vulnerabilities 3. Performance issues 4. Readability improvements

Be specific — quote the line of code and explain the issue and the fix.

* *[paste code]* *"

Study Question Template:

"Generate [N] practice questions from the following notes. Mix difficulty levels: [easy/medium/hard split]. Include answers with explanations.

Topic: [topic] Exam board/level: [details]

Notes: [paste notes]"

Content Brief Template:

"Write [content type] about [topic]. Audience: [who] Tone: [formal/casual/etc.] Length: [word count] Must include: [requirements] Must avoid: [exclusions] Format: [structure]"

Why Templates Work

• They ensure you do not forget important context or constraints
• They produce more consistent results over time
• They save time on repeated tasks
• They can be shared with teams to standardise AI use
• They improve over time as you refine them based on results

---

Step 3: Develop a Quality-Checking Habit

Every piece of AI output should go through a quality check before you use it. The intensity of the check should match the stakes.

The Quality Check Spectrum

Stakes	Check Level	Example
Low	Quick scan for obvious errors	Personal brainstorming, rough notes
Medium	Careful review for accuracy and tone	Work emails, blog posts, study materials
High	Thorough fact-checking and expert review	Published articles, academic submissions, production code
Critical	Multiple independent verifications	Medical information, legal documents, financial analysis

The 30-Second Quality Check

For medium-stakes output, a quick quality check covers: