First, a reminder of the funding rules
Under Circular Memorandum 221/2025, each approved publicly funded school receives a one-off grant of HK$500,000, to be used across the 2025/26 to 2027/28 school years, with a final use date of 31 August 2028 and any unspent balance returned to the Quality Education Fund. If a school fails to complete the deliverables, the EDB can ask for the whole grant back. Budget planning therefore has two bottom lines: the money must be spent well within the deadline, and every deliverable must be completed.
Working the budget back from the five deliverables
Rather than fixing ratios first and then finding things to spend on, a safer method is to budget for each deliverable. What software licences do three subjects of AI-assisted teaching need? How much co-planning and development time do six teaching demonstrations need (supply cover or outside support)? What equipment and recording do three open lessons and three sharing sessions need? What resources or outside services do two student activities need? Once every deliverable is budgeted, what is left is the flexible reserve.
This has an advantage: the budget sheet is itself a progress plan. When the interim report has to account for how the funding was used and how the work is progressing, the two line up naturally.
Typical allocation ratios for reference
Every school's situation differs, so the following are only the reasonable ranges commonly seen in the market, to check your own draft against: teaching and marking software licences around 30 to 40 per cent (over three years); teacher professional development and on-site support around 20 to 30 per cent; student activities and literacy resources around 10 to 20 per cent; supporting hardware around 10 to 20 per cent; and the remaining 10 per cent or so as a flexible reserve for needs that only surface in the second and third years, such as adding a subject or changing tools.
Watch for two signals. Hardware above 30 per cent usually means the focus has drifted away from AI-assisted teaching itself; the circular is about teaching applications, not equipment upgrades. If all the software sits with one supplier, reread the EDB's reminder about packaged solutions.
Planning spending across the years
- 2025/26 (around 30 to 40 per cent): the first batch of software licences, core teacher training, and the first student activity. The aim is to get three subjects genuinely using it.
- 2026/27 (around 30 to 40 per cent): renewals and adjustments to the tool mix based on trial results, support for open lessons, the second student activity, and deeper training.
- 2027/28 (around 20 to 30 per cent): the final year's renewals, plus organising results and support for the report. Do not leave a large purchase to this year, as a new tool no longer has time to produce results worth reporting.
The three most common budget mistakes
First, spending it all in year one: drawn in by a 'one-stop solution', the school signs a big three-year contract at the outset. From then on it loses the power to bargain or change direction, and when it finds in the second year that a tool does not fit, there is no money left to swap it.
Second, favouring hardware over use: turning most of the budget into tablets or displays. Equipment on its own is not evidence for any deliverable, and the report has no substance to write about 'AI-assisted teaching'.
Third, keeping no reserve: the budget sheet comes to exactly HK$500,000. There are always unplanned needs during the work, perhaps a subject wants to expand after a trial, perhaps a tool raises its price partway through. Keep around 10 per cent so the plan has room to adjust.