Ivo Nikolov All free resources

free template1 page · 2 min read

Three rules for any AI in your business

Write them down before you connect an AI agent to anything. Fill in this page once, and change it when the job changes.

Download the PDF

The template

Rule 1. What it can touch

Write down every app and folder it can open. Read-only wherever you can. Start with one inbox, and add the next app when you trust it.

Rule 2. What it asks first

Anything that leaves the business waits for your OK: a message, a post, a payment. It can draft all day. You still press send.

Rule 3. Where it writes it down

Every action goes in one log, with the time. You read it once a day. If something goes wrong, the log tells you where to look.

Fill it in with AI

Paste this into ChatGPT or Claude before you connect anything. Fill in the job.

Paste this

I want an AI agent to do one job for my business: [the job]. Help me write its three rules: what it can open, what it must ask me before it sends, posts or pays, and where it logs what it did. Ask one question at a time. Then give me the rules on one page.

Why write them down now

As of October 2026, OpenAI has told more than 100 organizations that its agents may have gone into their systems without permission. Meta's new Muse for Small Business (US and Canada only) promises that nothing sends, posts or spends without your approval. Your agent needs the same limits, on paper, before it starts.

The three rules at work

This is an example log of how my AI operators work. They're already working for estate agencies in Spain.

Time What happened
09:02A known contact writes. It stays quiet.
09:15A new enquiry comes in. It drafts a reply.
09:16The draft waits for the broker's OK.
09:20Sent, and written to the log.

What this page can't do

A page of rules doesn't stop an agent by itself. Set the same limits in the tool's own settings, where it lets you, and read the log. This page isn't legal advice on contracts or data.