I'm annoyed that the AI world is drowning in prompt collections, ten thousand magic phrases and no real implementation. Here's what actually works, and it fits on a sticky note. Every task you want AI to do reduces to two questions: What is the output I want? And where is the data to produce it? Answer both clearly and it can almost always do the job.
“I'm annoyed that AI is full of these prompts and no real stuff to implement.”
Brief it like an employee
- Output: not "help me with vendors" but "a spreadsheet of every vendor, grouped by country, with contact info and a one-line description." If you can't describe the deliverable, no prompt will save you.
- Data: where would a new employee find this? Your inbox, your call transcripts, a folder of documents. Point the AI at the same place, context is the difference between an answer and a hallucination.
- Then the third question: is this repeatable? If yes, schedule it, and a task becomes a system that runs without being asked.
Claire's own numbers
The framework, step by step
- 1
Output
not "help me with vendors" but "a spreadsheet of every vendor, grouped by country, with contact info and a one-line description." If you can't describ
- 2
Data
where would a new employee find this? Your inbox, your call transcripts, a folder of documents. Point the AI at the same place — context is the differ
- 3
Then the third question
is this repeatable? If yes, schedule it — and a task becomes a system that runs without being asked.
Claire's own framework from this essay, in the order she teaches it.
Source: Claire B. Soares, from her own travel business and live trainings.
Start on paper
My honest advice, from someone who lives in these tools: go back to paper first. Write down the repeatable tasks you touch every week, the Monday report, the inquiry replies, the follow-up list. That handwritten inventory is your automation roadmap; work down it one task at a time, asking the two questions of each. The advisors who fail at AI are collecting prompts. The ones who win are running a list.
Common Questions
What is the simplest framework for using AI at work?
Answer two questions: what is the output you want, and where is the data that produces it. Then ask a third: is this repeatable enough to schedule?
Why do prompt libraries fail?
Because they solve for phrasing rather than for delegation. Describing the deliverable and pointing at the source data matters far more than magic wording.
How should you start automating a business?
On paper. Write down the repeatable tasks you touch every week, then work down that list one task at a time.
Written By
Claire B. Soares
6× Condé Nast Top Travel Specialist, Certified AI Consultant, and founder of Travel AI University. $3.4B+ in enterprise travel software sold to Google, IBM, Boeing, the federal government, and the DoD; $16M+ in luxury travel sales with Up in the Air Life.
About Claire
