AI for After-School Programmes: Seven Admin Jobs and Their Risks

An after-school programme can hand AI the repeat writing and organising: reminders, rota drafts, parent messages, policy wording, flyers and payment chasers. A person must keep every decision about a child, every safeguarding judgement and every number that reaches a parent. Children's personal details stay out of public AI tools.

The split that works is simple. AI drafts from templates and made-up examples; you fill in real names inside the system you already use for registration, and you read every message before it leaves. The seven jobs below are ordered from lowest risk to highest, with what each one needs, an example, and the point where it goes wrong.

Follow me on Instagram@sagnikteaches

The line between drafting and deciding

In a small club the manager often works part-time with one deputy, and loses the same hour or two each week to rewriting near-identical messages. That is the work AI is good at: producing a first draft of text whose facts you already hold. It is poor at things that depend on knowing a particular child, a particular parent or what happened at pickup yesterday.

Connect on LinkedInSagnik Bhattacharya

So each job below gets a risk label. Low means the text contains nothing personal and a wrong sentence is merely embarrassing. Medium means real figures or dates appear and a mistake costs money or confuses families. High means the content touches a child's welfare, and AI may only tidy wording you wrote yourself.

Subscribe on YouTube@sagnikteaches

Job 1: flyers, posters and term-start notices

This is the safest place to begin because nothing personal is involved. A flyer for a new term needs a headline, three or four lines of detail and a call to action, and AI will give you five versions in under a minute. Your job is to pick one and correct the facts.

Give it the facts as a block, then ask for tone and length. A prompt that works:

Write three short flyer versions for a school-pick-up club.
Facts: Mon-Thu 3:15-5:45pm, ages 5-11, $14 a session or $48 for four,
starts 6 January, snack included, 24 places, collection from the
classroom door. Tone: warm, plain, no exclamation marks. Max 70 words each.
Do not add any claim I have not listed.

The last line matters. Left alone, an assistant will happily add "qualified, experienced staff" or "award-winning activities". A typical illustrative output reads: "Pick-up club, Monday to Thursday, 3:15 to 5:45pm. A calm space for homework, games and a proper snack. $14 a session or $48 for four. Twenty-four places, so book early." What you would fix: "calm" is a promise you may not be able to keep on a wet Friday, and "book early" is pressure you may not want. Swap those, check the pricing against your price list, and stop.

Photographs are the catch. If you use pictures of children on a flyer, check your photo permissions for each child first. AI image tools can create generic illustrations of children playing, which sidestep the question but should be labelled as illustrations if parents might think they are real.

Job 2: reminders for sessions, trips and payment dates

Reminders are the highest-volume job in most clubs. A trip form due Friday, a half-term closure, a payment date, an early finish: each needs saying two or three times in slightly different words. Ask AI for a reminder template with clear slots, then fill the slots from your calendar. If your booking system sends reminders itself, use AI only to improve the wording of its default messages. The booking system tutorial covers how to set that up.

Here is a before and after. The before: "Reminder - the form for the museum trip needs to be in by Fri otherwise cant guarantee place." The after, drafted by AI from your rough note and checked by you: "Hello [first name], a quick reminder that the museum trip permission form is due this Friday, 12 March. We can only hold a place for your child once we have it. You can return the paper copy in the book bag or reply to this message with a photo of the signed form."

The after version has a date, a consequence and two ways to respond. It also carries a risk: AI will cheerfully invent a date if you leave one out. Always put the date in the prompt, and check it in the output against the actual calendar. A reminder that pairs "Friday" with a date that falls on a Thursday does more damage than no reminder at all.

Keep a small "approved wording" document of six or seven reminder types. Once the wording is right, you stop asking AI each time and just paste and edit. That also makes your messages consistent for staff who send them when you are off sick.

Job 3: policies and the parent handbook

Late collection, illness, behaviour, photographs, complaints, lost property, drop-off arrangements. Clubs need these written down, and most are written once and left to go stale. AI is good at turning a rough list of your actual rules into plain English that a parent will read.

The trap is asking it to invent the rules. Give it your practice instead:

Here are our current rules for late collection, in my own words:
- grace period 10 minutes, then $1 a minute
- after 30 minutes we call the emergency contact
- repeated lateness: a conversation with me, then a written warning
Turn this into a clear policy for parents, under 200 words, no legal
language. Use only these rules. List anything you think is missing as
questions for me at the end, separate from the policy.

The final instruction is the useful bit: you want the gaps listed as questions, not quietly filled with assumptions. Illustrative output might ask: "What happens if neither the parent nor the emergency contact answers?" That is a good question and one only you can answer, probably with your insurer or the body that registers you.

Any policy that deals with safeguarding, supervision or medicine must be checked by whoever is responsible for those in your setting before it goes to parents. A fee clause may also need a legal check, because rules on late-collection charges vary. AI gives you a draft that is faster to review; it does not make the policy valid.

Job 4: rotas, cover and room plans

Rota drafting is where AI saves real time for a club with six to twelve staff, if you give it your rules in plain words. Describe the ratio you work to, who is vetted for which activity, who cannot work Wednesdays, and the sessions that need two people for a trip or a swimming slot. Ask for two or three options and a note of which rule each option strains.

Illustrative case: a club with eight children's sessions a week, five part-time staff, a ratio of one adult to every ten children and a rule that nobody works alone. A prompt listing each person's available days and the number of children booked per session returns a grid. A common error is that the grid looks right but counts a person who is available but not yet cleared for unsupervised work. The assistant does not know your vetting status unless you tell it, and even then it can miss one line in a long list.

The check you should run every time: for each session, count adults present and children booked, compare against your ratio, and confirm no lone-working slot exists. That takes four minutes per week and catches what the assistant gets wrong. Staff names and availability are personal data, though lower risk than children's information; use first names or initials in the prompt if you use a public chatbot, and keep the master rota in your own system.

For a club that already uses a booking tool, ask the vendor whether it can export a headcount per session. That gives you a better input than a typed number, and it keeps the child-level detail inside the booking tool.

Job 5: the registration form and its consent wording

A registration form collects the most sensitive information a club holds: medical needs, allergies, who may collect, emergency contacts, photograph consent. AI should write the form's structure and wording, but should never see a completed form. The separate tutorial on building the registration form with AI goes through the questions in detail.

What to ask for: a clear question list, grouped by purpose, with a reason beside each field. A usable instruction is "list every field a registration form needs for a pick-up club, and for each one say why we need it and whether it is optional." The reasons help you spot a field you do not actually need, and collecting less is the best protection you have. If the answer to "why do we need their home address" is "we do not", delete the field.

Illustrative test: a manager with a thirty-child club drafts a form, then asks AI to read it as a parent who is anxious about data and list every question that would make them hesitate. The output flags "Why do you ask for a family doctor's name, and who sees it?" That is a prompt to add one line of explanation under the field. It is a fair use of AI because it works on the blank form, not on anyone's answers.

After the form is live, the answers belong in your registration system or a locked spreadsheet with limited access, not in a chat window.

Job 6: payment reminders and billing chasers

Chasing money is the job managers dislike most, which makes it a good candidate for drafting. AI writes a firm, friendly, three-stage sequence in a few seconds: a gentle nudge at three days overdue, a clearer message at ten days, a final notice with your next step at twenty-one.

The main worked example, with figures. Say a club has 40 families, $48 a month typical, and on average six pay late each month. Chasing by hand takes about 8 minutes per family including writing, checking the account and sending: about 48 minutes a month. With three saved templates and a spreadsheet of who owes what, the time per family drops to roughly 2 minutes, so about 12 minutes a month. That saves around 36 minutes monthly. Modest, but it also removes the awkward rewording each time, which is the part that tends to get put off.

The check that matters is amount and family. Never ask AI to calculate what a family owes from attendance records. Do the sum yourself or in your booking tool, then paste only the amount into the template. A wrong figure in a chaser creates a dispute; a wrong tone creates only a mild annoyance. Also remove any chaser from the queue if you know a family is in difficulty. A friendly template cannot know that a parent is going through a hard month, and a person can.

A softened line to keep: "Hello [first name], our records show $48 is outstanding for February. If you have already paid, please ignore this and tell us the date so I can check. If something has changed on your side, just reply; I am happy to talk it through."

Job 7: messages to parents about a child

This is the high-risk job, and the honest answer is mostly no. A message about a bump in the playground, a behaviour concern or a child who seemed upset has to carry facts as you saw them, in your voice. A polished AI draft can drift from what happened, add reassurance you did not mean, or sound identical to every other message and read as cold.

The acceptable use is narrow: tidying wording you have already written. Write the facts yourself in plain words, with no names, then ask for a gentler version of the same message. Add the child's name back only when you paste it into your own messaging system. Illustrative sequence: you write "[child] fell off the low climbing frame at 4:20, grazed left knee, cleaned and plaster on, said they were fine, I wrote it in the accident book." Asking for "a calm, clear note to a parent, using only those facts" gives you a draft you can adjust in 30 seconds.

Some messages should not touch AI at all: anything about suspected harm, a child's disclosure, an allegation about a member of staff or a medical emergency. Those follow your safeguarding procedure, with the designated person leading. Writing them in a chatbot first is the wrong order of events.

What children's data means in practice

The working rule is to keep identifiable detail inside systems you chose, and send AI templates and made-up examples. Name, photo, date of birth, school, address, medical condition, family circumstances and anything about behaviour are all identifying when combined. Removing the name is not enough if the rest still points to one child in a club of thirty.

Practical settings that help, with the labels as they read in November 2026. On a consumer account, switch off training on your chats: ChatGPT calls the switch "Improve the model for everyone" under Data controls, Claude calls it Model Improvement in its privacy settings, and in Gemini it is the Keep Activity setting. Better still, use a business plan such as ChatGPT Business, Claude Team or Gemini in Google Workspace, where the vendor states it does not train on your content by default. And keep any shared Project to templates only. These reduce risk but do not make it safe to paste a child's file. The nursery tutorial on children's data sets out the same boundary with more examples, and it applies equally to a club.

Three questions to ask before any new tool touches child data: where is it stored, who at the vendor can read it, and how do I delete it? If you cannot get a plain answer, the tool stays on templates only. Your own data-protection rules, the body that registers your club and your insurer have the final say. Tell parents in your privacy notice what tools you use and for what.

A one-afternoon plan for a small club

Do not try all seven at once. In the first hour, write your six reminder templates and a flyer, because they are low risk and visible. In the second hour, turn your late-collection and illness rules into a one-page policy and send it to whoever is responsible for approval. In the third, test a rota prompt on last term's real week, using initials, and compare with what you actually ran. Put billing chasers in week two, after you have checked the amounts in your booking tool.

Keep a one-page log: what you asked AI to do, who checked it, what you changed. After a month the log shows which jobs saved time and which created more checking than they saved. If you would rather map this with someone, that is what my AI implementation consultation covers.

Three numbers to track over a half-term

Measure three things over a half-term. Minutes spent on admin per week, before and after. Errors caught at your read-through (a wrong date, an invented claim), because a rising count means the prompt needs tightening. And parent replies that ask "what do you mean?", which tell you the wording is unclear whatever the draft looked like. If admin time has not fallen by the end of the half-term, look at the jobs you picked before blaming the tool; try the reminder and chaser templates first, because that is where the repetition sits.

Questions club managers ask before trying this

Can I paste a child's medical or allergy details into ChatGPT?

No. Keep allergy, medical, safeguarding and pickup-permission details in your registration system. Ask AI to write the wording of a form or a rule, then fill in real details yourself, inside the system you already control and have agreed with parents.

Do I need parents' permission to use AI for club admin?

If AI only sees templates and anonymous examples, there is nothing personal to consent to. If any tool will process names, photos or messages about children, say so in your privacy notice and check your own data-protection rules first.

Which job should a small club automate first?

Start with parent reminders for sessions, trips and payment dates. They are repetitive, easy to check, and contain no sensitive detail. A good first week is one reminder template plus a manager who reads each draft before it goes.

How long does it take to set these seven jobs up?

Most managers can have the reminder, flyer and policy drafts working in one afternoon. The registration form and rota take a little longer because of testing. Billing chasers are quick to draft but need a careful read of every amount.

Further reads

• How Nurseries Use AI to Write Daily Reports for Parents — How daily parent reports are drafted without exposing a child's details.
• Parents' Evening Booking System: Slots, Reminders and Clashes With AI — Slots, reminders and clashes when many parents need booking.
• Daycare Late Pickup Policy: Can AI Word the Fees Fairly? — Late-collection fees worded fairly, with AI as the drafter.
• AI Tools for Preschool Owners: Parent Updates and Daily Reports — Parent updates written with AI, from the preschool side.
• Can AI Plan a Dance Recital? Schedules, Costumes and Parent Emails — A worked example of event logistics and parent emails.
• Sports Club Code of Conduct: Player and Parent Versions With AI — Behaviour policies written for children and for parents.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OpenAI Data Controls FAQ and business data page, Anthropic privacy centre (Model Improvement) and Claude pricing page, Google Gemini Apps Privacy Hub and Workspace generative AI privacy hub, all checked 4 November 2026.