AI Implementation for Education in Australia: A Starter Plan

Spend days 1 to 30 choosing two jobs and writing down which personal information each touches, days 31 to 60 piloting those jobs with one approved tool and a human check, and days 61 to 90 measuring and handing over. Check the Privacy Act first: the Australian Privacy Principles apply to personal information you put into an AI tool.

Summaries of government frameworks are easy to find. What a school or education business actually needs is a calendar, an owner for each step and a short list of things it must not do. This plan assumes a small organisation: an independent school of a few hundred students, a tutoring company with a dozen tutors, a language school or a vocational training provider. Government schools should treat it as a way to organise work that their department's approved-tools list will govern.

Follow me on Instagram@sagnikteaches

The 90 days on one page

The plan is deliberately narrow. Two jobs, one tool and a short pilot teach more than a school-wide rollout, and they leave a clear record if a parent, a board member or the regulator asks what you did.

Connect on LinkedInSagnik Bhattacharya

Days 1 to 20: choosing the jobs worth doing first

Ask every staff member to list the five tasks that take most of their week and are mostly writing: parent emails, lesson materials, report comments, newsletters, enrolment replies, timetable cover notes, fee reminders. Put them in a table and score each from 1 to 5 on four things: hours a week it takes, how repetitive it is, how much personal information it holds, and what happens if the draft is wrong. Multiply the first two, then subtract the last two.

Subscribe on YouTube@sagnikteaches

The winner here is enrolment replies (5 x 5 = 25, minus 2 and 2). Worksheet drafting is a good second job because it can use invented or generic examples with no student data at all. Learning-support plans score negative, which is correct: they hold sensitive information and an error is costly, so they stay human-written for now. Pick the highest scorer plus the safest one, so your pilot has one job that tests the data rules and one that tests the time saving.

Check the tasks that look harmless

Writing a quiz for Year 9 maths holds no personal information. Writing "a progress note for [student name], who has an attention difficulty and is behind in fractions" holds sensitive information, and the second version is the one a busy teacher types at 9pm. Mark every task by what staff will realistically paste in, not by what the task is called.

The privacy principles that matter here

The Office of the Australian Information Commissioner (OAIC) published guidance on commercially available AI products on 21 October 2024 and updated it in January 2025. Its core point is that privacy obligations apply to any personal information you enter into an AI system and to output that contains personal information. For an education business that translates into five of the Australian Privacy Principles (APPs), each with one practical question.

• APP 3 (collection). If a tool generates or infers personal information about a student, that is a collection. Is it reasonably necessary for what you do? A tool that guesses a child's learning difficulty from their writing probably is not.
• APP 5 (notice). Do parents know you use AI to handle their child's information? Add one sentence to your enrolment form and privacy policy.
• APP 6 (use and disclosure). Are you using the information only for the reason you collected it? Pasting enrolment data into a chatbot to draft a marketing email is a different purpose.
• APP 10 (accuracy). Have you taken reasonable steps to make sure what the tool produced is accurate? A teacher reading every comment before it is sent is that step.
• APP 11 (security). Where does the data go, who can see it and how long is it kept? Get the answer from the vendor in writing.

The OAIC also recommends, as best practice, not entering personal information, and particularly sensitive information, into publicly available generative AI tools. Student health, disability and behaviour information is sensitive. In practice that means free consumer chatbots get invented or anonymised examples only, and anything with a real name goes through a tool your organisation has approved.

Are you covered at all?

The Privacy Act generally does not apply to a small business with annual turnover of $3 million or less, but there are exceptions. The OAIC covers health service providers whatever their turnover, defines them as providing services in relation to physical, emotional, psychological and mental health, and lists child care centres, private schools and private tertiary educational institutions among its examples. An independent school or an early learning service should therefore assume the principles apply. A tutoring company with a counsellor on staff, or a business that provides services under a Commonwealth contract, should check rather than assume. Even when exempt, parents compare you with schools that are covered, so following the principles is a sensible default. Ask a privacy adviser for anything that turns on a legal reading.

State and department rules to check before a pilot

In June 2025 Education Ministers endorsed a 2024 review of the Australian Framework for Generative AI in Schools, which sits behind the rules each jurisdiction and school sector writes for itself. Read the framework once for its six principles (teaching and learning; human and social wellbeing; transparency; fairness; accountability; and privacy, security and safety), but do not stop there, because the binding rules for your organisation are the ones from your own department or sector body.

1. Government schools: follow your state or territory education department's list of approved AI tools and its guidance for staff and students. Do not run a pilot with an unapproved tool, however good it looks.
2. Catholic and independent schools: ask your sector body or school association whether it has issued guidance, and read your board's policy on data and technology.
3. Training providers and private tutoring businesses: check your registration or funding conditions for data-handling clauses, and read any contract with a school you serve.
4. Early learning: check your service's own data rules before photographs, observations or child names go near any tool.

Write what you found, with the date and the person who checked it. That one page becomes the reply if anyone asks why you used a tool.

The one-page policy to write by day 30

A long policy nobody reads is worse than a short one people follow, and the longer tutorial on writing an AI usage policy covers the full version. For a 90-day pilot, five lines are enough. A filled-in example for a tutoring business:

1. Approved tool: [tool name], on the business account only.
2. Never enter: student names, health or learning-support details,
   photographs, or parent contact details.
3. Allowed: invented examples, topic outlines, anonymised work.
4. Every output is read by a tutor before it leaves the building.
5. Tell [named owner] the same day if something is pasted in by mistake.

Rule 5 matters more than it looks. People will make mistakes, and a blame-free way to say so the same day gives you time to ask the vendor to delete the data and to decide whether anyone needs to be told.

A shortlist of tools, with the prices that are published

Pick one general tool, not four. Prices below are the USD list prices the vendors publish; check the Australian-dollar amount you will actually be charged at checkout before you budget, because currency conversion and tax change the figure. Education discounts and nonprofit pricing exist for some vendors; ask.

If your staff already use Google accounts, read what Gemini includes for schools and what costs extra before you buy anything. If you are a one-person tutoring business, the tutor tool guide covers the individual plans.

Five questions to ask the vendor in writing

1. Is our content used to train your models by default, and can we turn that off?
2. Where is our data stored, and can we choose the region?
3. How long is it kept after we delete it or close the account?
4. Can we remove a staff member's access and their stored chats?
5. What is your process if there is a data breach involving our account?

Our privacy terms checklist shows where each answer lives in a vendor's policy.

Running the pilot, days 31 to 60

Choose three to six volunteers, not the whole staff. Give them the one-page policy and a prompt template for each job, and ask them to log every use. The first pilot tutorial covers the basics; here is what the log needs for an education business.

A prompt for enrolment replies, with the sample output

You are replying for a tutoring business. Use only the facts below.
Do not promise results or places. End with one question.
Facts: Year 7-10 maths and English; groups of up to 5; Tuesday and
Thursday 4-6pm; trial session free; fee $48 per session.
Parent message: "Hi, my daughter is in Year 8 and struggling with
algebra. Do you have space this term?"

Illustrative output: "Hi [first name], thanks for getting in touch. We run Year 8 maths in groups of up to five on Tuesdays and Thursdays from 4 to 6pm, and the first trial session is free. Sessions are $48 each. Which day suits you better?" What you fix: nothing about space, because the prompt gave the AI no availability data. A draft that says "we have places" would be a promise the tool cannot know. The tutor adds the actual availability before sending.

Example: report comments without names

A teacher drafts end-of-term comments using only a skill grid: "Fractions: secure. Algebra: developing. Effort: consistent." The tool returns a paragraph with a placeholder for the name, and the teacher adds the student's name and one personal observation in the school's own system. Time per comment falls from about eight minutes to about four, and no student data enters the tool.

Example: a mistake in the log

Illustration: in week two a tutor pasted a parent's email, including a child's medical condition, into the free chatbot on her phone. She reported it the same day under rule 5. The director asked the provider to delete the conversation, noted the incident, reminded the team that free tools get invented examples only, and moved that tutor's drafting to the approved account. A blame-free log is what made it possible.

Pilot measures and pass marks

Write the pass marks before the pilot starts. For a small education business these five are enough.

For enrolment replies, the sum looks like this. A tutor takes 7 minutes to write a reply and handles 20 a week, which is 140 minutes. With a draft and a check at 4 minutes each, it is 80 minutes, saving 60 minutes a week or about 52 hours over a 52-week year. If the tool costs $25 a month, that is $300 a year for roughly 52 hours: a cost of under $6 an hour saved, before counting the faster replies parents receive. Replace the figures with your own measurements.

Budget for the 90 days, in Australian dollars

Illustrative figures for a tutoring business with 12 tutors, valuing staff time at $60 an hour. Replace them with your own.

The pilot hours are mostly the normal work with a new method, so the true extra cost is lower. Be honest about which part is genuinely extra: the audit, checks, policy and hand-over total about $1,920 plus the subscriptions. That is the figure a board is approving.

Days 61 to 90: decide, then hand over

Compare the log against the pass marks. For each job choose keep, change or stop. A job that saves time but fails on accuracy gets one more fortnight with a tighter prompt. A job that cannot reach the pass marks stops, and you write down why so nobody re-runs the same experiment next year.

Hand-over means a running procedure. Write one page per kept job: the prompt template, the data rule, the checker, where outputs are stored and the review date. Name one owner, brief the full staff in a 30-minute session that begins with the incident example above, and book a three-month review. Handing over to a named person, with the policy and templates in a shared folder, is what separates a pilot from a habit. Getting staff to follow it day to day is a separate job, covered in the rolling-out tutorial listed below.

The 30-minute staff briefing

The briefing is where a pilot becomes a practice. Keep it to four parts and give each a time limit: five minutes on what the two jobs are and why those, ten on a live demonstration with invented data, ten on the data rules with the incident example, and five on who to tell and when. Hand out the one-page policy and the prompt templates at the end rather than the start, so people listen rather than read.

Example: a newsletter paragraph, before and after

Illustration: a language school wants a short term-start paragraph for parents. The first draft from the tool reads: "We are thrilled to announce an exciting new term packed with opportunities for every learner." The editor replaces it with the facts only the school knows: "Term 3 starts on Monday 20 July. Beginner Spanish moves to Wednesdays at 5pm, and the Thursday intermediate class has two places left." The AI draft supplied the structure and a polite closing; the human supplied everything a parent will act on. Add a line to your template that says "use only the facts provided" and the filler largely disappears.

What to leave until after the first 90 days

Do not start with student-facing chatbots, automated marking that affects grades, anything using photographs of children, or tools that predict behaviour or ability. They raise fairness, consent and accuracy questions that a first plan should not carry. Do not buy a second tool until the first has a named owner. The five workflows to automate first lists the next candidates.

If you would like the audit and scoring done with someone, that is what my AI implementation consultation covers.

Questions education owners ask before starting

Does the Privacy Act apply to a small tutoring business?

Not automatically. A business with annual turnover of $3 million or less is usually exempt, but not if it is a health service provider, and the OAIC's examples of those include child care centres and private schools. Check the OAIC's small business page, and follow the principles anyway because parents expect it.

Can teachers paste student work into a free chatbot?

The OAIC recommends not entering personal information, especially sensitive information, into publicly available generative AI tools. Use anonymised or invented examples in free tools, and keep named student data inside a tool your organisation has checked and approved.

Do I need a separate AI policy?

A one-page policy is enough to start. It should list approved tools, banned data types, who checks outputs and who to tell if something goes wrong. Add it to your existing privacy and acceptable-use documents rather than creating a long standalone document.

What does a first 90 days cost?

Mostly staff time. A small team can run a pilot in about 90 staff hours in total, much of it normal work done a new way, plus anywhere from $0 to about A$40 a seat a month for tools. The one-off cost that surprises people is the time spent writing and testing the privacy checks.

Should government schools follow this plan?

Government schools should follow their state or territory education department's rules on which AI tools are approved, which can differ from this plan. Use the steps as a way to organise the work, and check approvals with your department before any pilot.

Further reads

• AI Implementation in Schools: A One-Term Plan for a Small School — A week-by-week version for one small school term.
• Why Your AI Pilot Stalled, and How to Get It Live — What to do when the pilot loses momentum in week five.
• AI for a Music School: Lesson Notes and Parent Emails Worked Through — Lesson notes and parent emails worked through for one school.
• How to Pilot AI in Shadow Mode Before Customers See It — Run the pilot with a human checking every output first.
• Does a Five-Person Business Really Need an AI Policy? — Whether a small team needs a written policy at all.
• AI for Education Consultants: Seven Jobs, Proposals to Reports — Seven jobs for advisers who sell to schools.
• How to Roll Out an AI Policy So Staff Actually Follow It — Getting staff to follow the policy after the hand-over.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OAIC guidance on privacy and the use of commercially available AI products (21 October 2024); OAIC small business privacy page; Department of Education, Australian Framework for Generative AI in Schools and its resource poster (2024 review endorsed June 2025); Microsoft Learn Copilot overview, Claude pricing page and an archived copy of OpenAI's business pricing page, checked 4 November 2026.