How Many Employees to Include in Your First AI Cohort (and Who to Leave Behind)

The founder asks: how many employees to take into the first AI cohort, and who to leave on the job so operations don't stall. Take too many — the process grinds to a halt. Take too few — you lack the scale for automations that move the business. There's no universal formula, but there is a logic used by founders who've already walked this path.

What Criteria to Use When Selecting for the First Cohort

The first cohort isn't about who wants to learn. It's about who can quickly deliver a result visible on the P&L. So we build selection around three questions.

First: does the employee have a task that repeats daily or weekly and takes more than 30 minutes? If yes — they're a candidate. Examples: gathering data from multiple sources for a daily report, transcribing calls, generating standard letters, filling CRM templates.

Second: can the task be described in words without code? If the employee says: «I take this file, compare it with that, fix errors, send it there» — that's a good signal. If it requires intuition, tactful contact, or creative solutions — we leave them for the next cohort.

Third: does this task block others? If one employee spends two hours so another can start their work — automation frees not just their time, but the flow.

During selection, we don't look at role or seniority. We look at the task. It could be a junior manager spending a day compiling tables, or a senior specialist doing the same. Both are candidates.

Which Employees to Leave on Duty During Training

Training the cohort takes two weeks, two hours per session plus homework — roughly five working hours per person per cycle. If you pull too many people from operations — you create a bottleneck.

Keep in operations:

• Those responsible for money flows: sales, invoicing, collections.
• Those who are the only ones who can perform a specific operation: for example, only one technician knows how to adjust the line for a new product.
• Those who work with clients in real time: support, delivery, order intake.

This doesn't mean they'll never be trained. They go into the second cohort after the first has streamlined processes and freed up mental space for new automations.

Checklist: Is Your Company Ready for the First Cohort

✅ Owner or CEO ready to dedicate 2 hours per week to review results and remove blockers. ✅ Clear definition of 3–5 tasks the team calls «time-sinks». ✅ Access to data and tools where this work happens (CRM, spreadsheets, email). ✅ No one promises training replaces hiring — it's about doing more with what you have.

If even one point is missing — start with a free org chart to see where routine is concentrated: https://course.aiadvisoryboard.me/uk/orgchart?utm_source=blog&utm_medium=article_body&utm_campaign=orgchart

How to Measure Success of the First Cohort Not by Satisfaction, But by Automations

Training satisfaction is a weak indicator. It doesn't correlate with whether AI brings money. So we measure something else.

Each participant must launch their first micro-automation by the second session. This isn't a demo — it's a tool running on company data doing what was agreed.

By the end of the cohort (four sessions) there should be at least three working automations on priority tasks the company selected before launch. They require no constant oversight, run on company tools, and are company property.

This is the result that impacts the P&L: hours freed from routine, errors reduced, faster response times.

If after the cohort there are no such automations — the training failed, regardless of how many people said it «was useful».

How This Works on Our Side

In our corporate program we work with two groups of 20 people each at a fixed price. Each participant describes their task in words — AI writes the code. By the end of the cohort each has their own working automation on a priority company process. First-month support is included in the price; code and automations are company property. https://course.aiadvisoryboard.me/corporate?utm_source=blog&utm_medium=article_body&utm_campaign=corporate

FAQ

Can you start with just one person? Technically yes, but the result will be local. One person won't shift the flow — there's no scale for comparison, idea exchange, or building a system, not just isolated scripts.

Are technical skills required? No. Participants describe task logic in words — e.g. «if column A in the spreadsheet is greater than B, then send an email». AI turns that into code.

What if half the team wants to go and the other half doesn't? Take based on task, not desire. If a person lacks repetitive, formalizable work — their place is in the next cohort, even if they're asking to join.

Can you spread training over weeks instead of a two-week intensive? Possible, but results will be slower. Intensity creates focus: people don't switch between tasks and training, they strive for results within a fixed timeframe.

If after the first cohort nothing changed — what to do? Review whether automations are being used in work. If not — possibly the task was chosen poorly (too complex or not priority) or there's no owner willingness to retire old reports/processes.

Conclusion

Your first AI training cohort isn't about covering as many people as possible. It's about selecting the right ones who'll quickly deliver results that free up time and money. Start with 8–12 people with clearly defined tasks, leave on the job those who protect revenue, and measure success not by satisfaction, but by launched automations. Tomorrow take the first step: run your company's free org chart to see where routine is concentrated — and ready for AI.