As a travel agency owner, I use AI to draft email responses. Some go to clients, some go to vendors, and some go to people reaching out to me about speaking gigs.
The thing people tell me most about their own AI drafts is that they can't trust them. The information isn't accurate, or the email doesn't sound human. Usually both. There are two ways to fix that, and they work best together.
“Harriet goes through and she changes the text and the content to make sure that it sounds like an actual human and not a bot.”
Make the draft loop
When AI writes an email, or any kind of correspondence, you don't want that first draft going straight out. You want it to loop. The draft goes through two checks, in order, before it ever reaches a person.
How it works
The loop every AI email goes through
- AI drafts
- Vicki verifiesare the facts right?
- Harriet humanizesdoes it sound like you?
- You approve and send
Something wrong? Back to the draft
Claire's own numbers
The loop every AI email goes through
- 1
Draft
AI writes the email or correspondence
- 2
Vicki the Verifier
a sub-agent checks what's actually in the message
- 3
Harriet the Humanizer
rewrites it to sound like a human, not a bot
- 4
Send
only after both have passed it
Draft, verify, then humanize, in that order, before anything is sent.
Source: Claire B. Soares, from her own travel business and live trainings.
One: Vicki the Verifier
Vicki is a sub-agent, and her only job is to verify what's actually in the message you're about to send. Is the information in this email correct? She checks the draft before anyone works on how it sounds, because there's no point polishing a sentence that's wrong.
Two: Harriet the Humanizer
If Vicki finds something wrong, the draft goes back to be fixed, and then it goes through Vicki again.
After Vicki has approved the draft as good to go, it goes to Harriet. Harriet goes through and changes the text and the content so it sounds like an actual human wrote it and not a bot.
To give you a feel for it, an example of the kind of instruction you might give each one (an illustration, not my own prompt):
- Example for Vicki: check every fact in this draft against what I gave you, and flag anything you can't confirm before it moves on.
- Example for Harriet: rewrite this approved draft so it sounds like me talking to this person, without changing any of the facts.
Why the order matters
Verify first, then humanize. If you humanize first, you end up with a warm, friendly email that's still wrong. Getting the facts settled first means Harriet is only working on how it sounds.
A note before you try this
Test the loop on one email before you trust it with your inbox, and read what comes out next to what you would have written. Vicki can only check facts against what you give her, so give her the booking, the itinerary or the email thread she's checking against. And keep yourself as the last step. Nothing goes out until you've read it and pressed send.
Set it up with me
If you came here from Instagram, comment Loop Me on the reel and I'll send you the prompt to set this up yourself. And if you want to build your first AI employee with me live, join my free AI Masterclass.
Common Questions
Why do AI-drafted emails sound robotic or contain mistakes?
Because the first draft goes out unchecked. Nothing verifies the facts and nothing rewrites the tone. Claire fixes this with a loop: the draft goes to a verifier sub-agent that checks the content, then to a humanizer that rewrites it so it sounds like a real person wrote it.
What is Vicki the Verifier?
Vicki the Verifier is the name Claire gives a sub-agent whose only job is to check what's actually in a drafted email before it's sent. She reviews the information in the message, and the draft only moves on to the humanizing step once Vicki has approved it as good to go.
Should I verify or humanize an AI draft first?
Verify first. If you humanize first, you can end up with a warm, natural email that still says the wrong thing. Settling the facts before the rewrite means the humanizer, Harriet in Claire's setup, only works on how the message sounds and never on what it claims.
Written By
Claire B. Soares
Claire B. Soares is the founder of Travel AI University, a 6x Condé Nast Traveler Top Travel Specialist, and a Certified AI Consultant with a B.S. in Computer Science. A top 10% enterprise sales executive, she sold $3.4B+ in enterprise software to Google, Microsoft, IBM, Boeing, the federal government, and the DoD, then built $16M+ in luxury travel sales.
About Claire
