6.3: Edit Before Send
In the last section, you learned how to approve AI output with a simple checkbox. That's great! But sometimes you want to go one step further: What if you could edit the AI's draft before sending it?
Imagine this: AI writes 90% of your email in perfect grammar and structure, but you want to:
- Add a personal touch
- Fix a small detail
- Adjust the tone slightly
- Include specific context only you know
Instead of starting from scratch OR accepting the draft as-is, you get the best of both worlds: AI does the heavy lifting, you add the finishing touches!
Let's build this powerful pattern.
6.3.1: Why Edit-Before-Send Is Powerful
Think about how you'd normally work with AI:
Option 1: AI writes, you copy-paste-edit
- ❌ AI drafts in one place
- ❌ You copy to your email client
- ❌ Edit there
- ❌ Send manually
- ❌ Lots of switching between tools
Option 2: Approve-only workflow (what you learned in 6.2)
- ✅ AI drafts automatically
- ✅ You approve with one click
- ❌ But you can't make small tweaks
- ❌ It's all-or-nothing
Option 3: Edit-Before-Send (what you'll learn now!)
- ✅ AI drafts automatically
- ✅ You can edit right in the table
- ✅ Make any changes you want
- ✅ Approve when ready
- ✅ Your edited version sends automatically
💡 The Best of Everything
Edit-Before-Send gives you:
- 🚀 Speed — AI writes the first draft
- ✏️ Control — You refine it perfectly
- 🎯 Quality — Combines AI efficiency with human insight
- 🔄 Workflow — Everything happens in one place
It's like having an assistant who writes the draft, and you just polish it before hitting send!
6.3.2: What We're Building
Before we dive into the setup, let's be crystal clear about the workflow:
The Flow:
- Customer sends email → Creates new row
- AI reads and drafts response → Fills "AI Draft" column
- AI Draft auto-fills "Final Email" column ← The magic!
- You click the cell and edit the text ← Your polish!
- You check "Ready to Send?" when satisfied
- Your edited version sends (not the AI draft!)
Edit Before Send: Complete Workflow
🔑 The Key Pattern
- 1️⃣ AI generates the draft (fast, automated)
- 2️⃣ You edit and refine it (quick, just polish)
- 3️⃣ You approve when ready (quality control)
- 4️⃣ Your version sends automatically (no copy-paste!)
Key insight: The "Final Email" column starts with the AI's draft, but you can change anything. It's editable! When you approve, your edited version is what gets sent.
6.3.3: Step 1 - Add an Editable Text Column
First, you need a column where the AI draft appears AND where you can edit it.
Here's how:
1. Open Your Table
- Go to your "Support Tickets" table (or whatever table you're using)
- You should already have an "AI Draft" column that generates responses
2. Add a New Text Column
- Click "+ Add Column"
- Choose type: "Text (Long)" — This allows multi-line text
- Column name:
final_email - Column label: "Final Email"
- Click "Add Column"
3. Position It Correctly The column order should be:
- Customer Message (input)
- AI Draft (generates response)
- Final Email (editable version) ← Your new column!
- Ready to Send? (approval checkbox)
- Send Email (sends the final version)
💡 Why This Column?
This "Final Email" column is special because:
- ✅ It's a text column (you can edit it)
- ✅ It starts with the AI draft (pre-filled)
- ✅ You can change anything (full control)
- ✅ It's what actually gets sent (not the AI draft!)
Think of it as your "working copy" that begins with AI's suggestion.
Final Email (Editable)
This column is pre-filled with the AI draft, but you can edit it before sending
Configuration
6.3.4: Step 2 - Pre-Fill with AI Draft
Now comes the clever part: Configure "Final Email" to automatically copy the AI draft when it's ready.
Configuration:
1. Edit Your "Final Email" Column
- Click the column menu (three dots)
- Choose "Edit Column"
2. Set the Default Value
- Look for "Default Value" or "Pre-fill from" setting
- Select: "AI Draft" column
- This tells the system: "When AI Draft completes, copy its content here"
3. Enable Manual Editing
- Make sure "Allow manual editing" is enabled (should be by default for text columns)
- This lets you click cells and edit them
4. Save the Configuration
- Click "Save"
Configure Column: Pre-fill with AI Draft
Step 2Pre-fill Configuration (The Magic!)
- ✓ When a new row is created, this column starts empty
- ✓ Once "AI Draft" completes, its content automatically fills this cell
- ✓ You can then click the cell and edit the text
- ✓ Your edited version is what gets sent
Users can click any cell in this column to edit its contents
💡 💡 How Pre-Fill Works
What happens automatically:
- Customer email arrives → Creates row
- AI Draft column runs → Generates response
- As soon as AI Draft completes, its content copies into Final Email
- You see the draft text in Final Email column
- You can now click that cell and edit it!
You don't have to do anything — the copying is automatic! You only interact when you want to edit.
6.3.5: Step 3 - Configure Send to Use Edited Version
Critical step: Make sure your "Send Email" column uses the edited version, not the original AI draft!
Here's how:
1. Edit Your "Send Email" Column
- Click the column menu on your "Send Email" column
- Choose "Edit Column"
2. Check the Email Body Configuration
- Look for "Email Body" or "Message" field
- It should reference: "Final Email" (the editable one)
- NOT "AI Draft" (the original)
3. Verify Dependencies
- Make sure "Send Email" waits for:
- ✅ "Ready to Send?" checkbox (approval)
- ✅ "Final Email" has content
- This ensures your edited version is ready before sending
4. Save
- Click "Save"
Configure Send Email Column
Step 3⚠️ Critical: Use the Edited Version
Make sure your "Send Email" column uses Final Email (the editable one), NOT "AI Draft" (the original AI output).
References the column with customer's email address
What Gets Sent:
→ Ignores your edits
→ Your changes are wasted
→ Includes all your changes
→ Perfect control!
Waits for approval before sending
💡 ⚠️ Common Mistake
Double-check this!
If "Send Email" uses "AI Draft" instead of "Final Email":
- ❌ Your edits get ignored
- ❌ The original AI version sends
- ❌ You wasted time editing!
Make absolutely sure the send column references "Final Email" (your editable column), not "AI Draft"!
6.3.6: Step 4 - Test the Complete Workflow
Okay, everything's configured! Let's test it to make sure the whole flow works.
Follow these steps:
1. Create a Test Email
- Add a new row manually
- Enter a customer question in the message column
- Example: "What are your business hours?"
2. Watch AI Generate Draft
- The "AI Draft" column runs automatically
- It creates a response
- Look at "Final Email" — the draft should appear there too!
3. Edit the Final Email
- Click on the cell in the "Final Email" column
- The text becomes editable
- Make some changes:
- Change the greeting
- Add a personal note
- Adjust a phrase
- Save your changes (click outside the cell or press Save)
4. Approve It
- Check the "Ready to Send?" checkbox
- Watch the "Send Email" column activate
5. Verify What Was Sent
- Check that the email sent
- Verify: Did your edited version send, or the original AI draft?
- You should see your changes in the sent message!
💡 🎉 Success!
If it worked, you now have:
- ✅ AI automatically drafting emails
- ✅ You editing them right in the table
- ✅ Your polished version sending automatically
- ✅ No copy-paste, no switching tools!
This is powerful! You've combined AI speed with human precision in one seamless workflow.
6.3.7: Real-World Usage Tips
Now that it works, here are tips for using this workflow effectively in your daily work:
Tip 1: Quick Edits Are Fastest
You don't need to rewrite everything! The most efficient use is:
- ✅ Quick fixes: Typos, wrong names, incorrect details
- ✅ Personal touches: Adding "Hope you're doing well!", mentioning previous conversations
- ✅ Tone adjustments: Making it warmer, more formal, more casual
- ✅ Context additions: Info only you know ("As we discussed on the phone yesterday...")
Most of the time, you're editing 5-10% of the draft, not rewriting it from scratch!
Tip 2: Create Edit Templates in Your Mind
After using this for a while, you'll notice patterns:
- AI always forgets to mention the customer's name → You add it
- AI is too formal → You add friendly phrases
- AI doesn't reference order numbers → You add them
Pro move: Give AI better instructions to reduce these repetitive edits! Improve your "AI Draft" column's prompt.
Tip 3: Batch Your Editing Time
Instead of editing emails one-by-one throughout the day:
Morning routine (10 minutes):
- Filter for "AI Draft complete + Final Email not edited"
- Edit 10-15 drafts in one session
- Approve them all
- Done for the morning!
This is WAY faster than constant interruptions.
Tip 4: Train the AI by Editing
Here's a secret: Your edits actually help improve the AI over time!
When you consistently edit certain things:
- The system notices patterns
- You can update the AI's instructions
- Future drafts get better
- You edit less and less
Over weeks, the AI learns your style!
💡 The Learning Loop
Week 1: Edit 50% of drafts (lots of tweaks) Week 4: Edit 30% of drafts (AI is learning) Week 12: Edit 10% of drafts (AI nails your style!)
The workflow gets faster over time as the AI adapts to your preferences!
6.3.8: Advanced Pattern - Handle Uncertain Cases
Here's a powerful advanced pattern: What if the AI could flag when it's uncertain and ask for your help?
The Scenario:
Sometimes AI isn't confident about its answer:
- Customer asks about an edge case
- Knowledge base has conflicting information
- The question is ambiguous
Instead of guessing, the AI can detect this and flag it for human review!
AI Uncertainty Detection
Needs ReviewAI Confidence: Low (62%)
The AI is uncertain about this response. It's flagged for human review.
"Can I return a custom-printed item if I don't like the color?"
"Thank you for your question. Our return policy typically allows returns within 30 days. However, custom items may have different rules. I recommend checking with our returns department to confirm the specific policy for custom-printed items."
Why This Was Flagged:
- • AI found 2 different policies in knowledge base
- • One says "no returns on custom items"
- • Another says "returns within 14 days with 20% restocking fee"
- • Confidence score: 62% (below 80% threshold)
- • Automatically marked for human review
"Thanks for asking! For custom-printed items, we accept returns within 14 days if there's a printing defect or error on our part. If you simply changed your mind about the color, we can't accept the return since the item was custom-made for you.
However, before you order, we're happy to send you a digital proof to make sure you love the colors! Just reply to this email if you'd like to see a proof first."
🧠 AI Learns from Your Correction
When you provide the correct answer, the AI system notes:
- • The correct policy interpretation
- • Which knowledge base article is more current
- • The better way to phrase this response
Over time, the AI becomes more confident and accurate on similar questions!
How to implement this:
1. Add a Confidence Score Column
- Before "Final Email", add a column that checks AI confidence
- Many AI systems return a confidence score (0-100%)
- You can also write custom logic: "If knowledge base returns conflicting results, flag as uncertain"
2. Add a Conditional Flag
- Add a column: "Needs Review" (checkbox)
- Condition: "If confidence < 80%, check this box"
- Or: "If AI detects multiple conflicting answers, check this box"
3. Filter for These Cases
- Create a saved view: "Uncertain Responses"
- Shows only rows where "Needs Review" is checked
- You prioritize these for editing
4. Provide Corrections
- When you edit these uncertain cases, you're teaching the AI
- Document the correct answer
- Update your knowledge base
- Future similar questions become confident!
💡 Smart Escalation
This pattern is brilliant because:
- ✅ AI handles questions it's confident about (90%)
- ✅ Flags edge cases for human expertise (10%)
- ✅ You focus on the hard problems
- ✅ System gets smarter over time
It's like having a junior assistant who knows when to ask for help!
6.3.9: Edit-Before-Send vs. Approve-Only
When should you use edit-before-send vs. simple approval?
Use Edit-Before-Send when:
- ✅ Responses need personalization
- ✅ Context varies by customer
- ✅ Tone needs adjusting per situation
- ✅ You want to add specific details
- ✅ You're still training the AI
Use Approve-Only when:
- ✅ Responses are highly standardized
- ✅ AI accuracy is consistently high (95%+)
- ✅ Speed is more important than customization
- ✅ Volume is very high
- ✅ AI has been trained well
💡 Start with Edit, Evolve to Approve
Recommended approach:
- First month: Use edit-before-send, refine AI prompts
- Second month: As AI improves, edit less
- Third month: Switch certain message types to approve-only
- Ongoing: Keep edit-before-send for complex cases, approve-only for routine ones
You can have BOTH patterns in the same table for different scenarios!
6.3.10: Troubleshooting Common Issues
Running into problems? Here's how to fix them:
Issue 1: Final Email Column Stays Empty
Cause: The pre-fill isn't configured correctly
Fix:
- Edit "Final Email" column
- Check "Default Value" setting
- Make sure it references "AI Draft"
- Save and test with a new row
Issue 2: My Edits Aren't Saving
Cause: Column might not be set to allow manual editing
Fix:
- Edit "Final Email" column
- Look for "Allow manual editing" or "User can edit" setting
- Make sure it's enabled
- Save and try editing again
Issue 3: Original AI Draft Sent Instead of My Edit
Cause: "Send Email" is using wrong column
Fix:
- Edit "Send Email" column
- Check the "Body" or "Message" configuration
- Change from "AI Draft" to "Final Email"
- Save and test again
Issue 4: Can't Find Where to Edit
Cause: Trying to edit the wrong column
Solution:
- Don't try to edit "AI Draft" (that's generated automatically)
- Click on "Final Email" column instead
- That's the editable one!
What You've Learned
Incredible work! You've mastered the edit-before-send pattern. Let's recap what you can now do:
✅ Create editable text columns that pre-fill with AI output ✅ Configure pre-fill to copy AI drafts automatically ✅ Edit content directly in the table with easy cell editing ✅ Send edited versions instead of original AI drafts ✅ Combine AI speed with human precision for perfect results ✅ Handle uncertain cases by flagging them for review ✅ Choose the right pattern for each use case
The Power You Now Have
Think about all the workflows you can build:
Customer Support: AI drafts, you personalize, auto-send Sales Emails: AI researches and writes, you add personal touch, auto-send Content Creation: AI writes articles, you edit and polish, auto-publish Social Media: AI creates posts, you refine tone, auto-schedule Documentation: AI writes first drafts, you review and enhance, auto-deploy
You've unlocked the perfect balance: AI does 90% of the work, you add 10% of your expertise, and together you produce 100% quality results!
What's Next?
You've now completed Part 6: Human Oversight! You know how to:
- Identify when humans need to review (6.1)
- Set up approval workflows (6.2)
- Edit AI output before sending (6.3)
In the next part (Part 7: Organizing Your Data), you'll learn how to:
- Filter your data to see exactly what you need
- Save favorite views for quick access
- Create dashboards to visualize your automation
These skills will help you manage large-scale automations efficiently. Ready to get organized? Let's go!
Previous: 6.2 Setting Up Approval WorkflowsNext: 7.1 Filtering Your Conversations