The "First Draft, Not Final Draft" Mindset
The Core Problem
Professionals fall into two traps with AI-generated content:
Trap 1: The Blind Acceptor
They take whatever AI produces and use it verbatim. Copy, paste, send. The result? Generic content that sounds like everyone else, potential factual errors, and a voice that doesn't match their brand or personality. Worse, they're outsourcing their judgment along with the task.
Trap 2: The Perfectionist Rejecter
They try AI once, get mediocre output, and declare it useless. "See? It doesn't understand my industry." They abandon a powerful tool because they expected magic on the first try.
Both traps share the same root cause: misunderstanding what AI is actually good at.
AI is not a replacement for your expertise, judgment, or voice. It's a drafting accelerator. It gets you from blank page to rough draft in minutes instead of hours. What you do with that draft is where your value lives.
The Mindset Shift
Think of AI like a capable but new junior colleague:
| Junior Colleague | You (The Senior) |
|---|---|
| Produces first drafts quickly | Knows what "good" looks like |
| Works from templates and patterns | Understands context and nuance |
| Needs direction and feedback | Applies judgment and expertise |
| Might miss industry specifics | Catches errors and fills gaps |
| Available 24/7, never tired | Responsible for final quality |
You wouldn't hand a client deliverable to a new hire and send it without review. You also wouldn't refuse to delegate anything because their first draft wasn't perfect.
The Mantra
"AI gets me to the starting line faster. I still run the race."
The 60% Principle
Here's the mental model that changes everything: AI gets you to 60% done.
That 60% is the scaffolding—structure, ideas, initial language. Your job is the final 40%:
- Accuracy and fact-checking
- Voice and personality
- Industry-specific nuance
- Strategic judgment
- The "this is actually me" factor
Without AI: 0% → 100% (you do everything)
With AI correctly: 0% → 60% (AI drafts) → 100% (you refine)
The second path isn't about doing less. It's about spending your time on the high-value 40% instead of grinding through the low-value 60%.
The Refinement Framework
Once you have an AI draft, apply this four-step review:
1. Fact Check (Accuracy)
AI confidently states things that may be wrong. Your first pass should catch:
- Statistics or data points (verify the source)
- Quotes or attributions (confirm they're real)
- Technical claims (check against your knowledge)
- Dates, names, and specific details
Red Flag
If something sounds too perfect or too specific, that's often where hallucinations hide.
2. Voice Check (Authenticity)
Ask yourself:
- Does this sound like me or like a generic professional?
- Are there phrases I'd never actually use?
- Is the tone right for my audience and my relationship with them?
Common AI tells to edit out:
- "In today's fast-paced world..."
- "It's important to note that..."
- "Let's dive in..."
- Excessive hedge words ("arguably," "potentially," "it could be said")
- Overly formal phrasing where casual fits better
3. Value Check (Substance)
Does the draft actually say something useful, or is it eloquent fluff?
- Cut filler sentences that don't advance the point
- Add your unique perspective, data, or examples
- Remove anything that's technically true but unhelpfully vague
4. Action Check (Clarity)
If this communication requires a response or action:
- Is the ask crystal clear?
- Is there only one way to interpret what's needed?
- Have I made it easy to say yes (or respond appropriately)?
Before and After: The Refinement in Action
Scenario: Writing a Client Update Email
AI Draft (60%):
Subject: Project Update - Q4 Dashboard Initiative
Dear Client,
I hope this email finds you well. I wanted to provide you with an update
on the Q4 Dashboard Initiative project.
Our team has made significant progress over the past two weeks. We have
completed the initial data integration phase and are now moving into
the visualization development stage. The project remains on track for
the December 15th deadline.
There are a few items we need from your team to continue moving forward:
- Confirmation of the KPI priorities
- Access to the historical data warehouse
- Availability for a 30-minute review session next week
Please let me know if you have any questions or concerns.
Best regards
After Your Refinement (100%):
Subject: Dashboard on track—need 3 quick things by Friday
Hey Sarah,
Quick update: data integration is done, and we're building out the
visualizations now. Still on track for Dec 15.
Need three things from your side to keep momentum:
1. **KPI priorities** – Is revenue growth still #1, or has the board
shifted focus to retention? (This changes which charts get
prime real estate)
2. **Data warehouse access** – Credentials for Marcus on your IT
ticket #4521
3. **30 min next week** – Want to show you the revenue viz before
we build out the rest. Tuesday or Wednesday work?
If I have these by Friday, we're golden.
Talk soon,
[Name]
What Changed:
- Generic opening → direct and specific
- Vague "items we need" → numbered, specific asks with context
- Formal tone → matches existing relationship with client
- Added the "why" behind each ask (so Sarah can prioritize)
- Clear deadline (Friday) for action items
- Subject line that actually tells you something
Exercise 1: The Refinement Practice
Generate a draft using AI, then apply the four-step refinement process.
Your Task: Use AI to draft an email declining a meeting request from a colleague.
Step 1: Generate the Draft
Prompt to use:
Write a professional email declining a meeting request from a colleague.
The meeting is a "brainstorm session" for a project I'm not involved in.
I need to decline politely but clearly, without leaving the door open for
negotiation. Keep it brief.
Step 2: Apply the Refinement Framework
| Check | Questions to Ask | Your Edits |
|---|---|---|
| Fact Check | N/A for this type of email | |
| Voice Check | Does it sound like me? Too formal? Too casual? | |
| Value Check | Am I saying what I actually mean? Is there fluff? | |
| Action Check | Is my "no" clear? Will they understand? |
Step 3: Write Your Final Version
[Paste your refined version here]
Exercise 2: The 60/40 Split
This exercise helps you see where AI adds value versus where you add value.
Your Task: Choose a real piece of content you need to create this week. Use AI to draft it, then track your edits.
The Tracking Template
| Section/Paragraph | AI Contributed | I Changed | Why I Changed It |
|---|---|---|---|
| Opening | |||
| Main point 1 | |||
| Main point 2 | |||
| Main point 3 | |||
| Closing/CTA |
After completing the exercise, answer:
- What percentage of the final version was AI vs. you?
- Where did AI save you the most time?
- Where did you add the most value?
- What would you prompt differently next time?
Exercise 3: The Voice Injection
Most AI output sounds generic because it lacks your personality. This exercise fixes that.
Part A: Collect Your "Voice Samples"
Find 3 pieces of your writing that represent your authentic voice:
- An email you're proud of
- A document or report that sounds like you
- A message that got a positive response
Part B: Extract Your Voice
Paste your samples into AI and use this prompt:
Analyze these three writing samples from the same author.
Describe their writing style, including:
- Typical sentence length and structure
- Vocabulary preferences (formal vs casual, jargon vs plain language)
- How they open and close communications
- Any distinctive phrases or patterns
- Their approach to humor, directness, and warmth
Part C: Apply Your Voice
Save the AI's analysis. Next time you use AI to draft something, include:
Write in this style: [paste your voice analysis]
Compare the output to a generic prompt. Notice the difference.
Quick Reference: The Refinement Checklist
Use this for every AI draft before sending:
□ FACT CHECK
□ Statistics verified?
□ Names/dates correct?
□ Claims I can stand behind?
□ VOICE CHECK
□ Sounds like me, not generic AI?
□ Removed AI-isms ("In today's world...", "It's important to note...")?
□ Tone matches audience and relationship?
□ VALUE CHECK
□ Actually says something useful?
□ My unique perspective included?
□ Filler and fluff removed?
□ ACTION CHECK
□ Clear ask or next step?
□ One way to interpret what's needed?
□ Easy to respond to?
The "Edit Tells" Cheat Sheet
When reviewing AI drafts, watch for these patterns that signal generic output:
| AI Tell | Why It's a Problem | Fix |
|---|---|---|
| "In today's fast-paced world..." | Cliché filler | Cut entirely or get specific |
| "It's important to note that..." | Hedge that adds nothing | Just state the thing |
| "This helps to ensure that..." | Passive, wordy | Use active voice |
| Long lists without prioritization | Everything seems equal | Rank or cut |
| Absence of specific examples | Sounds theoretical | Add your real data/stories |
| Perfect parallel structure | Too polished, feels robotic | Vary sentence patterns |
| No contractions | Overly formal | Use can't, don't, we'll |
Key Takeaways
- AI produces drafts, not deliverables. Your job is the refinement that makes it actually good and actually you.
- The 60% Principle changes the game. AI handles scaffolding; you handle substance, voice, and judgment.
- Four checks catch everything. Fact Check, Voice Check, Value Check, Action Check—run every draft through this filter.
- Generic AI tells are fixable. Learn to spot "In today's fast-paced world" energy and edit it out.
- Your value is the final 40%. That's where expertise, personality, and judgment live. Don't outsource it.
Next Steps
- [ ] Generate an AI draft for a real task this week
- [ ] Apply the 4-step refinement framework before using it
- [ ] Track your edits using the 60/40 Split exercise
- [ ] Create your "voice analysis" doc for future prompts
- [ ] Save the refinement checklist somewhere accessible