How Many AI Employees Do I Need? The Honest Answer

Airun Company · August 25, 2026 · 5 min read

How many AI employees do I need is the wrong question, and the right question is how many tasks do I want off my plate. The honest answer to the headline is: fewer than you think, usually two or three, and one before that. I built seven because my company has seven distinct work lanes, not because seven is a magic number. The count follows the work, and the work follows the lanes, and the lanes are what you should count.

Why the number is smaller than you expect

Most businesses have one or two tasks that eat the week, and everything else is noise. One employee handling the newsletter and the social recap, one handling support and follow ups, and suddenly the week has a shape it did not have before. The people with ten AI employees are usually running ten half working lanes, and the people with two are running two that actually deliver. Two working beats ten aspirational.

Count lanes, not headcount

• List the tasks you do every week
• Group them by output: drafts, replies, summaries, follow ups
• Each group is a lane, and each lane is one employee
• One lane, one rules file, one review
• Add a lane only when the last one runs clean

The lane counting takes an hour and it answers the question better than any formula. A lane is a repeatable task with a clear output and a schedule. If you can write the sentence every Monday, we draft the newsletter, you have a lane. If you cannot write the sentence, you do not have a lane yet, you have a wish.

The first one is the hard one

The first employee takes the longest because you are building the files, the review habit, and the trust all at once. Expect a weekend of setup and two weeks of review before it runs clean. Every employee after that takes half the setup, because the templates exist and the habit is formed. The first one is also the one that teaches you everything, so the slowness is an investment, not a delay.

The two employee setup

For most solo businesses, two employees cover the week: one for content and marketing, one for support and admin. The content employee drafts posts, newsletters, and pages from your notes. The support employee answers the inbox, sends follow ups, and files the paperwork. Both run on rules files, both get a weekly review, and both free up the hours that were going to busywork. That setup is the free starter kit plus an afternoon of adapting it.

When to add the third

You add a third employee when one lane outgrows its file, not when you feel behind. The sign is simple: a lane's review is taking more than thirty minutes a week, or the lane has two jobs that do not share a skill. That is the moment to split the lane into two employees with two files. The split takes an hour and the review time drops back down, because each file got simpler.

The ceiling is your review time

Every employee needs a weekly review, and your review time is the real limit. At twenty minutes per employee, seven employees is about two and a half hours of review a week, which is sustainable. The ceiling hits when review time crowds out the work itself, and that is the signal to automate the review, not to add more employees. The org chart in my company follows this rule, and the structure is in the book if you want to copy the shape.

The honest bottom line

Start with one. Prove it. Add the second. Prove it. Count lanes, not headcount, and scale one at a time. If you want the starting files for the first two employees, they are free in the starter kit, and the full lane map is in the book. The answer to how many you need will come from your own week, and it will be smaller than you think.

The part-time employees nobody counts

The honest math also includes the part-time lanes: the task that takes an hour a week does not need a full employee, it needs a half lane or a shared lane. A single employee can run two small lanes with two files and two review slots, and the count should reflect the review time, not the task count. The businesses that overcount are the ones with ten files and no time to review them, and the ones that undercount are the ones still doing the work by hand. The lane math sorts both out.

The number that changes with the season

The count is not fixed, it breathes with the business: launch season adds a content lane, quiet season cuts it back, and the review time follows the lanes that are actually running. The system is designed for that breathing, because the files are portable and the lanes are divisible. The mistake is treating the employee count as a permanent org chart instead of a weekly decision. The weekly review is where the count gets rechecked, and the count that matches the work is the count that works.