AI Employee Training: The 3 Files That Turn a Bot Into Staff
AI employee training sounds like a course, but it is really three files and a habit. The files are the rules, the tasks, and the plan, and the habit is the weekly review. Chatbots come pretrained, and employees are made by the files you write around them. Here is the training system that turns a bot into staff.
File 1: The rules
- Who the employee is and what its lane is
- The tone to write in, with one good example
- What it should never do
- Who to escalate to and when
- What good output looks like, with one bad example
The rules file is the job description and the contract in one. It should take half an hour to write, and if it takes longer, the job is too vague. The examples matter most: one good and one bad example teach the tone faster than any paragraph of description.
File 2: The tasks
One week of tasks with owners, outputs, and due dates. The task list is what turns a chatbot into an employee: it gives the work a schedule instead of waiting to be asked. Keep it in plain text so it travels with you, and keep it small enough to review in one sitting.
File 3: The plan
Where the work is going and why. The plan file is the direction, and it is what keeps the employee from optimizing the wrong thing. One page: the goal, the current lane, the next lane, and the definition of done. The plan is the answer to the question what are we building, and it changes slowly.
The training week
- Day 1: write the three files
- Day 2: run the first real task
- Day 3: check the output and fix the rules
- Day 4: run the second task with the fixes
- Day 5: set the weekly review and log what changed
That is the entire onboarding, and it fits around your actual work instead of replacing it. The second employee takes half the time because the files already exist and you are just adapting them.
The review that trains
Thirty minutes a week, same day every time. Read everything produced, fix the two worst outputs in the rules, queue next week. The review is where the training actually happens, because every correction is a lesson the rules file records. Skip it for two weeks and the output drifts back to generic.
When training is done
Training is done when a week passes with only small corrections and the review takes ten minutes. That is the signal to widen the lane or add the next employee. The system I use is the same one in the book, and the three files are free in the starter kit, ready to adapt.
What good rules look like
Good rules are short, specific, and loaded with examples. Bad rules are long, abstract, and full of adjectives like high quality and professional, which the model cannot act on. The rewrite is the same idea in concrete terms: write like the example, never use words from the banned list, escalate anything about money. The concrete version works, and the abstract version produces the generic output everyone complains about.
The training rhythm after week one
After the first week, training happens in the review, not in a session: every correction is a rule update, and every rule update is a training step. The rhythm is what makes the employee better without any extra time, because the review was happening anyway. The log records the rhythm, and the log is what shows the trajectory: corrections shrinking, review shrinking, output improving.
How to know the files are working
The files are working when you can onboard a new employee from them in an evening, and when a skipped week of review is immediately visible in the output. Both are tests, and both are easy to run. The files are working when the system survives the test, and the system is the asset, not the platform. The book's training section covers the tests, and the starter kit has the three files ready to adapt.
The difference between training and prompting
Prompting is asking for a one off answer. Training is building the files and the habit that make the answers consistent. The difference shows in the output: prompted answers vary, trained output converges on your voice. The files are the training, the review is the reinforcement, and the log is the record. The three files and the habit are the whole training program, and the program is the book's shortest chapter on purpose.
Set it up the right way
The book walks through the full system: 4 files, the org chart, the failure modes, and a 30-day blueprint. $29, plain English, 30-day refund.
Get the book, $29