Lesson 2: Build It - Your Email Writing Workflow
Estimated Time: 15 minutes
Learning Objectives:
- Break down your manual process into AI-friendly steps
- Build a prompt template you can reuse every time
- Test your workflow with three real examples
- Know when to use your workflow vs. writing from scratch
From manual process to AI workflow
In Lesson 1, you documented your current process for responding to customer emails. Maybe it looked something like this:
- Read the customer's email
- Figure out what they're asking
- Check if you've answered this before
- Draft a response
- Make it sound friendly and professional
- Add personalization (their name, reference their specific situation)
- Proofread
- Send
Here's what most people do wrong: They go to ChatGPT and type, "Write a response to this email."
And they get back something generic that doesn't match their voice, doesn't answer the question properly, and needs so much editing that they might as well have written it themselves.
The trick is turning your manual process into a reusable template that gives AI everything it needs to do the job right.
Let's build that template together.
The four-part prompt template
Every good workflow prompt has four parts. Think of it like a recipe - you need all the ingredients or the dish doesn't come out right.
Part 1: Context (Who you are and what you're doing)
Tell AI your role and the situation. This sets the tone and style.
Example:
"You are a customer support specialist for a small e-commerce company. You're responding to customer questions about orders, shipping, and returns."
Part 2: Task (What you need AI to do)
Be specific about the output. Don't say "help me" - say exactly what you want.
Example:
"Write a professional but friendly email response to the customer's question below."
Part 3: Format (How you want it structured)
Include things like length, structure, and any required elements.
Example:
"Keep it under 150 words. Use this structure: greeting, answer their question directly, offer additional help, friendly closer."
Part 4: Constraints (Rules AI must follow)
These are your non-negotiables. Tone, things to avoid, specific words to use or not use.
Example:
"Tone: warm and helpful, not corporate. Always use the customer's name. Never make promises about shipping times - just say we'll check on it. Sign off with 'Happy to help!' and your name."
Real example: Customer support email template
Let's put it all together. Here's a complete prompt template for responding to customer inquiries:
You are a customer support specialist for [Your Company Name], an online store that sells [your products]. You're known for being helpful, friendly, and responsive.
Write a professional but warm email response to the customer question below.
Structure:
- Greeting using their name
- Direct answer to their question (1-2 sentences)
- Any relevant additional info they should know
- Offer to help with anything else
- Sign off with "Happy to help, [Your Name]"
Rules:
- Keep it under 150 words
- Tone: friendly and conversational, like you're texting a friend who happens to be a customer
- Be specific - no vague "we'll look into it" unless you genuinely don't know
- If you don't have enough info to answer, ask a clarifying question
- Never use corporate language like "we apologize for any inconvenience" - just be human
Customer's email:
[PASTE THEIR EMAIL HERE]
Notice what this does:
- AI knows who you are and your brand voice
- It has a clear structure to follow
- It knows your tone rules
- It knows what NOT to do (no corporate speak)
This is your reusable template. You save this somewhere (Notes app, Google Doc, Notion) and every time you need to respond to an email, you paste it into ChatGPT and swap in the customer's email at the bottom.
Where to save your template
You need this template somewhere you can access it quickly. Here are three options:
Option 1: Text expansion tool (fastest)
If you use something like TextExpander, Alfred, or Espanso, save your template with a shortcut like ";;email" so you can trigger it with a few keystrokes.
Option 2: Notes app (simple)
Create a note called "AI Email Template" and pin it to the top. Copy-paste when needed. Works great on phone too.
Option 3: ChatGPT saved prompt (if you have ChatGPT Plus)
ChatGPT Plus lets you save custom instructions. You can make a custom GPT just for email responses with this template built in.
Pick whatever makes it easiest for you to actually use. The best system is the one you'll stick with.
Test it: Three real examples
Let's test your template with three different types of customer emails. This is how you know if your workflow actually works.
Test Case 1: Easy (straightforward question)
Customer email:
"Hi, I placed an order last week (order #4521) but haven't received a shipping notification yet. Can you check on this? Thanks, Sarah"
Your prompt:
[Paste your full template from above, with Sarah's email at the bottom]
AI output (example):
"Hi Sarah,
I just checked on order #4521 and it looks like it shipped out yesterday via USPS. You should receive a tracking email within a few hours (sometimes there's a delay in the system syncing). If you don't see it by end of day, let me know and I'll send you the tracking number directly.
Happy to help,
[Your name]"
Time to complete: About 30 seconds to paste template and customer email, 10 seconds for AI to generate, 20 seconds to review and send.
Total: About 1 minute. Compare that to your baseline from Lesson 1 (probably 8-10 minutes).
Test Case 2: Medium (needs slight customization)
Customer email:
"I ordered the blue medium t-shirt but received a red large. I definitely need the blue medium. What do I do now?"
Your prompt:
[Same template, paste this email at bottom]
AI output (example):
"Hi [Customer name],
Oh no, sorry about the mix-up! Here's what we'll do: I'll send you a prepaid return label for the red large, and I'm putting a blue medium in the mail today so you get it faster. You should have it by Thursday.
Just print the label when it arrives in your email (should be there in the next hour), slap it on the box, and drop it at any USPS location. You won't pay for return shipping.
Anything else you need, just let me know!
Happy to help,
[Your name]"
What you might tweak: Check your actual return policy and adjust the response if needed. Maybe you need to add "I've refunded your account for the difference" if sizes have different prices. The AI gives you 90% of the way there - you just personalize the last 10%.
Time to complete: 1 minute for AI + 30 seconds to customize = 1.5 minutes total.
Test Case 3: Tricky (missing info)
Customer email:
"Why hasn't my order arrived yet? I ordered it forever ago."
Your prompt:
[Same template, paste this email]
AI output (example):
"Hi [Customer name],
I want to help you track down your order! Can you send me your order number or the email address you used when you placed it? Once I have that, I can see exactly where it is and get you an update right away.
Happy to help,
[Your name]"
Why this works: The AI knows (from your constraints) that it can't give a vague answer. Since there's no order number, it asks for the info it needs. Professional, helpful, gets you what you need to actually solve the problem.
Time to complete: 1 minute.
When to use your workflow (and when not to)
Your workflow isn't for every email. Here's when to use it and when to write from scratch.
Use your workflow when:
- It's a question you've answered before
- The structure is similar to past emails
- You need to be professional and friendly but don't need to be creative
- You're responding to something routine (order status, basic info, confirmations)
Write from scratch when:
- It's a sensitive situation (customer is angry, you made a mistake)
- It's the first email to someone important (potential client, partnership)
- You need to be very personal or creative
- The situation is unique and doesn't fit your formula
Think of your workflow as a power tool. You don't use a drill for everything, but when you need to drill a hole, it's way faster than doing it by hand.
Same with AI workflows. They're for the repetitive stuff that follows a pattern. Use them there, and you'll save hours every week.
What you just learned
- How to turn your manual process into a four-part AI prompt template
- Where to save your template so you'll actually use it
- How to test your workflow with easy, medium, and tricky examples
- When to use your workflow vs. writing from scratch
You now have a working email response workflow. The template is built, tested, and ready to use.
Try It Challenge
Before moving to Lesson 3, do this:
- Take your prompt template and save it somewhere accessible
- Find three real emails from this week that you could have used it on
- Run each one through your template and see what AI gives you
- Time how long each one takes (be honest - include review time)
- Compare that to your baseline from Lesson 1
Write down the time difference. That's your proof that this works.
Example:
- Baseline: 9 minutes per email
- With AI workflow: 2 minutes per email
- Time saved: 7 minutes per email
- If you do this 15 times per week: 105 minutes saved every week
That's real. That's tangible. And that's just from one workflow.
Quiz
Q1: What are the four parts of a good workflow prompt template?
A) Introduction, body, conclusion, signature
B) Context, task, format, constraints
C) Question, answer, example, feedback
D) Title, description, steps, results
Answer: B
Context (who you are), task (what you need), format (how you want it), and constraints (rules to follow). These four parts give AI everything it needs to produce consistent, high-quality output every time.
Q2: You receive an angry email from a customer whose order was damaged in shipping. Should you use your AI workflow for this?
A) Yes, it'll save time
B) No, this is too sensitive and needs a personal touch
C) Yes, but edit it heavily afterward
D) Only if you're in a hurry
Answer: B
Sensitive situations need human judgment and genuine empathy. Your workflow is great for routine questions, but when someone is upset or the situation is delicate, write it yourself. AI can help you draft ideas, but you should own the final message completely.
Next lesson: You'll learn how to debug and improve your workflow when AI gets things wrong. We'll cover the three most common problems and exactly how to fix them.