What software the funding can buy

Under EDB Circular Memorandum 221/2025, the funding covers the costs of running AI-assisted teaching. Common categories include AI teaching and marking software, learning resources for student AI literacy, related hardware, teacher professional development, and on-site support services. The funding can be used until 31 August 2028, and any unspent balance must be returned to the Quality Education Fund.

What to buy should be worked backwards from the deliverables you have committed to, not forwards from a product catalogue. Schools have promised to run AI-assisted teaching in at least three subjects, to develop at least six teaching demonstrations, to hold three open lessons, three sharing sessions, and two student AI literacy activities. Every purchase should ideally map to at least one of these.

Four criteria for choosing AI education software

1. Fit with the curriculum

Does the product match the Hong Kong curriculum framework and your school's actual teaching progress? Content built around overseas curricula takes teachers extra time to adapt, and the real usage rate tends to stay low. Ask the supplier to demonstrate with a unit you are teaching right now, rather than watching an official demo clip.

2. Cantonese and Traditional Chinese support

Do the interface and content support Traditional Chinese? Does the voice function understand Cantonese? This matters especially in primary and junior secondary. A tool that does not speak the language ends up used only by a handful of English teachers, which makes the 'at least three subjects' promise hard to keep.

3. Student data policy

Where data is stored, how long it is kept, whether it is used to train models, and whether it meets the Personal Data (Privacy) Ordinance. Get written answers to these four questions before you sign. Our school AI tool privacy checklist goes into detail.

4. Three-year total cost

Work out the three-year total cost with August 2028 as the end point, including renewals, adding subjects, training, and the cost of exporting your data if you leave. A product with a cheap single-year price is not necessarily cheap over three years.

What to look at in each category of AI education software

AI marking platforms: the key measures are how well the marking scheme is localised and whether every mark comes with evidence. See our AI marking platform comparison guide for detail.

Student AI literacy resources: the key measure is classroom usability. Can a single lesson use it straight away? Beyond playing with AI, is there a step for checking outputs and discussing ethics? Does it leave behind student work you can write into the report?

Classroom management and administrative tools: the key measure is how quickly teachers pick it up. Tools like these affect a teacher's daily routine directly, and if they need long training before anyone can use them, the real adoption rate will be low.

General AI platform licences: buying a campus licence for a large platform gives you the most flexibility, but it needs teacher training and usage guidance to go with it, or the licence sits idle. A licence on its own does not amount to 'running AI-assisted teaching' in the report.

Avoiding the packaged-solution trap

In its clarification of 4 February 2026, the EDB gave schools a specific reminder to assess carefully whether packaged solutions suit the school's own needs. Some suppliers sell a whole bundle on the promise of 'handling all five deliverables in one stop', usually priced close to the funding ceiling. The risks are that the quality of individual components in the bundle varies, that the school is locked into a single supplier, and that ready-made demonstration examples may not reflect the school's real situation, which makes them hard to support fully in the report.

A safer order is to compare similar tools on the evidence first, trial on a small scale, confirm that teachers really will use them, and then buy in stages. Buying separately also keeps the school's bargaining power and its freedom to walk away.

A suggested procurement process

  • Work backwards from the five deliverables and list the gaps that no tool yet covers
  • Compare at least three options in each category, scoring them against the same set of criteria
  • Ask suppliers to demonstrate with your own subjects and year levels
  • Get the student data policy and the three-year total cost in writing
  • Have one subject panel trial it for a term and collect teacher feedback
  • Only decide on a whole-school or multi-subject purchase once the trial evidence is complete