The short answer
Train each office on its own tasks, in its own tools, with its own data in front of it, and leave behind two things: a written playbook of what's approved and how it's done, and one person in that office who owns it. A half day built on the admissions inbox changes how the admissions office works. A two-hour presentation on what AI is changes nothing. The difference is whether people leave the room having done their actual job a new way, with someone beside them.
Why the webinar didn't stick
Most schools have already tried the general session: a speaker, slides about what large language models are, a few impressive demos, a slide about ethics. Attendance was good and the next week looked exactly like the week before. The reason is that nobody left with a task they could now do differently. The demos were someone else's work, the tools weren't the ones on their desks, and nothing in the session answered the question every staff member actually has, which is "am I allowed to use this on this?"
The second reason is that AI's useful work in a school is specific and a little boring: inquiry replies, contract tables, reconciliation exceptions, reteaching material from real assessment data. None of it demos well in a ballroom. All of it transfers in a small room with the real inbox open.
Build the session from real work
Before the session, collect the tasks. Ask the office for the five things that take the most time each week and the three they dread. Bring the real material: last week's inquiries, the current contract folder, the board packet from last month. Then run the session as a working morning, not a class. Each task gets done twice, once the old way to agree on what "good" looks like, once with the tool, with the team deciding together what to keep, what to change, and where a person must approve before anything goes out.
Use the school's own accounts and tools, never a trainer's demo environment, so the setup that works on the day is the setup that's there on Monday. End with the office writing its own approval rule in one sentence and naming who approves. That sentence goes into the playbook and the policy.
One trained lead in every office
Training everyone a little produces a school where nobody is responsible. Training one person in each office thoroughly produces four people who are. The AI lead in admissions, the business office, the front office, and communications isn't the most technical person; it's the one who knows the work best and is willing to try things. They get extra time in the session, a direct line to whoever built the tools, and a short weekly job: notice the next task worth trying, test it on one case, and show the team.
This is the model that holds up after the trainer leaves. New colleagues learn from the lead, not from a recording. The next use case comes from the office that has the problem, not from a vendor's roadmap. And when the policy is reviewed, the leads are the ones who know what actually happened.
The playbook the school keeps
The session should produce a document the school owns: the approved tools and accounts, the tasks the office now does with AI and exactly how (the prompts, the templates, the steps), the approval rule and the approver's name, examples of good output from the school's own work, and a short list of what the office decided not to do and why. It lives where the office keeps its procedures, not in the trainer's drive, and the AI lead updates it. If a trainer leaves without a playbook, the school paid for a morning.
The first month after
The first month decides whether the training took. Three numbers tell the story: hours saved on the named tasks (the office's own estimate, written down weekly), reply or turnaround time where it applies, and errors caught, meaning the times the approval step changed or rejected what the AI produced. That last number should not be zero; if it is, nobody is reading the drafts. A short check-in at week two and week four, with the lead and the trainer, fixes the setup problems while they're still small and adds the second use case once the first is routine.
The mistakes to avoid
Training the whole school at once, so nobody gets depth. Starting with the hardest task because it would be the most impressive, when the easy one would have built trust. Running the session on a demo account, so Monday's tools don't match. Skipping the approval rule and discovering it when a draft reaches a parent. Treating the session as the end rather than the first of three touches. And measuring nothing, so the board asks what the training achieved and the honest answer is a feeling.
Training leadership
The head of school and the business officer need a different session from the offices: not how to use the tools, but how to judge them. What questions to ask a vendor, what the approval log should show, which numbers to expect in which month, what a reasonable first year looks like, and where the risks are. A leadership team that can read the approval log and the monthly numbers will make good calls about the next use case. One that only saw the demo will make calls based on the demo.
Want your offices trained on their own work?
EnrollmentGenius runs working sessions built on each office's real tasks, trains an AI lead in every office, and leaves the school with a playbook it owns, with the tools living in the school's own accounts.
Schedule a Call →Frequently asked questions
How should a private school train staff on AI?
Office by office, in a working session built on that office's real tasks and real tools, ending with a written approval rule, a playbook the school keeps, and one named lead per office who owns it afterward.
Who should be trained first?
The admissions office and the business office, because their clerical load is the largest and the gains show up within a month. Leadership gets a separate session on judging tools, vendors, and results.
What is an AI lead in a school office?
One person per office, usually the one who knows the work best, trained in more depth, with a weekly job of trying the next task, showing the team, and keeping the playbook current. They're the reason training holds after the trainer leaves.
What should the school be left with after AI training?
A playbook: approved tools and accounts, each task and exactly how it's done, the approval rule and approver, examples from the school's own work, and what the office chose not to do. Plus the trained lead and a follow-up schedule.
How do you measure whether AI training worked?
Hours saved on the named tasks, turnaround times where they apply, and the number of AI drafts the approval step changed or rejected. Review at two weeks and four weeks, then add the next use case.
Should teachers be trained the same way?
The method is the same, built on real tasks, but the tasks differ: reteaching material from the school's own standards and results, parent communication drafts, and planning. Anything that reaches a family or a student record still needs the teacher's approval before it goes out.