P1-S6 School planning: an annual AI review pack (a five-column review table, six evidence categories, three self-evaluation judgements, four points in a cycle)

• StageP1-S6
• SubjectSchool planning
• Learning strandthe fundamental operating principles of AI / human-AI communication and collaboration / reflection on the application of AI
• Official documentsBlueprint for Digital Education Development in Primary and Secondary Schools · AI Literacy Learning Framework for Primary and Secondary Schools · Guide to Using AI in Teaching in Primary and Secondary Schools · EDB Circular No. 156/2026 · EDB Circular No. 107/2026
• Time needed to prepareAbout 60 minutes to build the evidence file the first time; about 90 minutes for the annual review; plus one 60-minute review meeting

A five-column review table (school target, what was done this year, where the evidence is, result, next year's decision), six evidence categories, three self-evaluation judgements, and four points in one cycle. Internal review only: not appraisal, no ranking, no comparison.

• AudienceTeachers
• ThemeSchool planning

• EquipmentEditable Word file; the five-column table and the six evidence categories are one page each and photocopy on their own; the one-page meeting agenda can go into the room
• Version1.0
• Last updated2026-11-03

A school does a great deal of AI work in a year: two staff training sessions, one open lesson, one parents’ evening. Come July, asked how the year went, it holds a plan and no evidence, so the review meeting turns into a briefing, and next year’s plan repeats the same three targets.

What this solves

Three situations come up most often in school AI planning:

• A plan, and no evidence. The annual plan set three targets and a good deal was done, but nobody recorded a file name, a date and where it is kept for any of it. In July the meeting runs on memory.
• The review turns into a briefing. The coordinator speaks for fifteen minutes and everyone else listens; nobody checks what last year decided, so this year’s review has no connection to last year’s.
• A change of post starts from zero. A coordinator leaves or moves, the person taking over finds a plan and nothing else, and cannot tell what was done, what was not, or where the evidence is.

This page puts those three into one document: a five-column table, six evidence categories, three judgements and four points in a cycle.

Three things to say plainly first:

1. The name of this page (an annual AI review pack), the five-column table, the six evidence categories and the three judgements are a suggestion from this resource, not an Education Bureau requirement. What the Bureau requires is continuous improvement through a self-evaluation cycle, reviewing the effectiveness of the work in good time, and bringing digital education into the school development plan and/or the annual school plan. It prescribes no format, no columns and no wording for the judgement.
2. This page is for internal review only. It is not used for staff appraisal, it sets no ranking, and it compares the school with no other school. What is reviewed is the school’s year, not any colleague’s performance; nothing written here should become part of a personnel record.
3. Anything paid for from the “Smart Learning with AI” grant must still submit its mid-term and final reports under that scheme’s guidelines. This page lists the interface only: the figures, receipts and deadlines belong to that scheme’s guidelines and the school’s own accounts. It does not replace either report and sets up no second way of calculating anything.

Purpose

By the end of this pack a school will be able to:

1. Use the five-column table to turn a year’s work into five rows that can be checked: school target, what was done this year, where the evidence is, result, next year’s decision.
2. Use the six evidence categories to turn “we did it” into a document somebody can find, rather than a memory of an event.
3. Use the three judgements (in place / partly in place / not in place) to reach a conclusion that can be written down for every row, with one sentence of reason and one of next year’s approach where something is not in place.
4. Use the four points in the cycle to put the review back into the school calendar: collect in May and June, meet in July, write it into the school report and annual plan in September, and check last year’s decisions first in July of the following year.
5. Keep the review internal: no teacher’s or pupil’s name, no ranking, no borrowed outside indicators and no comparison with other schools.

The pack starts from nurturing the whole person, with technology as an aid, and takes as its values the four principles of technology for good, nurturing the whole person, equity and access, and forward-looking innovation. It is written around the three learning strands of AI literacy: the fundamental operating principles of AI, human-AI communication and collaboration, and reflection on the application of AI. The four attitude sets - critical discernment, timely and appropriate use, human-AI collaboration and technology for good - look, in an annual review, most concretely like a willingness to read last year’s page first in July rather than to open with this year’s new plan. It is a review document rather than a report card because its reader is the school’s next year: only what can be written in five columns gives next year a starting point.

How to use it

The file is a Word document in six parts: the five facts at the top, the five-column review table, the six evidence categories, the three self-evaluation judgements, the four points with a meeting agenda, and a one-page proofreading list. The curriculum lead should own it. Building the file the first time takes about 60 minutes; after that, the review takes about 90 minutes a year, plus one 60-minute meeting.

Step 1: the five-column table, and only five columns

The five are fixed; none is added and none removed. One row per school target; how many targets there are is for the school to decide, and there is no need to merge or split them for tidiness.

Three things worth saying about using them:

• Column 3 is fixed at three items so that somebody else can find it. With the file name, the date and the location written down, anyone holding this page can open the document; a screenshot breaks when a layout changes and a copy drifts from the original, so neither is used.
• Column 4 carries the judgement and one sentence. “Partly in place: the training was completed, but classroom use happened in two classes only” is fine. “Significant impact” and “an 80% achievement rate” are not: the first is a comment, and the second flattens a year into a number with no denominator.
• Column 5 has to be shorter than column 2. A decision is one sentence, not a paragraph; a decision written as a paragraph is one nobody re-reads next year.

Step 2: six evidence categories, and only six

The six are fixed. Each has one storage location and one post responsible for it, and the same document goes into one category only.

Three things to add under the evidence page:

• Six categories, and no seventh. Something that will not fit is usually not part of an annual review: it is a flyer from one event, or a colleague’s own notes.
• Categories 2 and 4 carry no names. Training records sessions, hours and numbers; pupils’ work records the year group and the number of pieces, with names covered on the sample. This is the same rule as “not appraisal”.
• Category 6 records the source and the deadline, and recalculates nothing. The mid-term and final reports go in under that scheme’s guidelines; this page records only that the report exists, when it is due and where it is kept.

Step 3: three self-evaluation judgements, and only three

The judgement goes in column 4, one per row, and no separate self-evaluation report is written.

Three boundaries:

• “Not in place” is not a failure, but it may not be left blank. One sentence of reason and one of next year’s approach is the whole requirement; “not enough time” with no next step means writing it again next year.
• The result is the judgement plus one sentence, nothing else. No achievement rate, no “significant impact”, no satisfaction score, and no converting the judgements into points and adding them up.
• The judgement takes the evidence as its floor. If something happened but no evidence can be found, the judgement is “not in place”, and the reason sentence says “not retained”. That is more honest than marking it higher, and it tells next year what to keep.

Step 4: one cycle, four points

The four points are fixed and their order does not change; the dates may move with the school calendar, but all four things happen.

The 60 minutes are best spent like this: 10 minutes reading last year’s table and the one line of record; 30 minutes working through this year’s five columns row by row; 15 minutes writing next year’s decisions; 5 minutes confirming that the six storage locations have not changed. The meeting notes are one page, attached after the table.

Step 5: internal use only

Three sentences go at the top of the review table:

• Not for staff appraisal. No teacher’s name appears, and no wording that points at a person, such as “led by the subject head of X”; where responsibility has to be written, write the post.
• No ranking. No ranking between subject teams, classes or years; the table is one row per target, not one row per unit.
• No comparison with other schools. No outside indicators and no other school’s figures; the standard is the target the school itself wrote down.

Step 6: the interfaces with school documents

Seven documents each do their own part. This page lists the interfaces, changes none of their wording and redoes none of their work:

• The whole-school AI risk register and incident process: incidents are recorded there; this page opens no incident column, and only notes in column 2 whether there was one.
• The AI planning continuity pack: a change of post hands over the state of seven documents; this page’s five-column table is part of the fifth of them, and the pack changes none of these columns.
• The scheme-of-work AI literacy mapping template: the mapping says which unit carries which literacy element; this page cites how far it is complete and does not redo the mapping.
• The student AI-voice survey and what-we-changed loop: the survey’s aggregate figures go into evidence category 4; this page designs no second survey and runs no second round.
• The external AI partner vetting pack: the vetting and after-event record of one outside activity live there; column 2 here says only that there was a talk and what kind it was.
• The AI budget and licensing worksheet: the three cost scenarios are worked out there; category 6 here records the source, the spending and the deadline, and recalculates no scenario.
• The whole-staff AI-literacy baseline self-assessment and CPD plan: the baseline figures and the training plan are there; column 2 and category 2 here cite the result.

On funding, one more sentence: anything paid for from the “Smart Learning with AI” grant must still submit its mid-term and final reports under that scheme’s guidelines. Category 6 records that the two reports exist, when they are due and where they are kept, and does not replace them.

Review checklist

Check once before the July meeting and once after:

• □ All five columns are present, in the order above, with no sixth added.
• □ One school target per row, and column 1 carries the original wording of the development plan or annual plan.
• □ Column 2 holds three to five concrete things per row, written as events rather than adjectives.
• □ Column 3 holds the three fixed items - file name, date, location - and there is no screenshot anywhere.
• □ Column 4 carries exactly one of the three judgements, with one sentence.
• □ Every row marked “not in place” carries one sentence of reason and one of next year’s approach.
• □ Nothing anywhere gives an achievement rate, a “significant impact” or a satisfaction score.
• □ All six evidence categories are present, in the order above, with no seventh added.
• □ Each category has a designated location and a post responsible, and no document sits in two categories.
• □ Categories 2 and 4 carry no teacher’s or pupil’s name, class or class number.
• □ Category 6 records the source, the spending and the deadline, and recalculates no amount.
• □ All four points are present in order, with an output at each: May to June, July, September, and the following July.
• □ The first step in the following July is to check last year’s column 5.
• □ The page says the review is internal, is not used for appraisal, sets no ranking and compares the school with no other school.
• □ The page says the name, the five-column table, the six categories and the three judgements are this resource’s suggestion, and gives what the Bureau does require - a self-evaluation cycle and reviewing the effectiveness of the work in good time.
• □ The page says the “Smart Learning with AI” grant still requires its mid-term and final reports, and that this page lists the interface only.
• □ The seven school-document interfaces are listed only, with no wording, column or count changed.

Files

Annual AI review pack: evidence file and self-evaluation (Word document, editable)

Seven pages, each usable on its own: the five facts at the top (stage, subject, learning strands, the official documents it maps to, time needed to prepare); the five-column review table (one page); the six evidence categories (one page); the three judgements (one page); the four points and the meeting agenda (one page); the interfaces with seven school documents (one page); and the proofreading list (one page). A school fills in its own targets, dates and locations; the file carries no box for a teacher’s or pupil’s name, class or class number.

Official documents

Every resource in this library is written to the AI Literacy Learning Framework for Primary and Secondary Schools and falls within the policy scope of the Blueprint for Digital Education Development in Primary and Secondary Schools.

This resource maps to:

• Blueprint for Digital Education Development in Primary and Secondary Schools (three-fold leaflet, p.2): its four principles are technology for good, nurturing the whole person, equity and access, and forward-looking innovation; this page puts the annual review under forward-looking innovation and nurturing the whole person - the review exists to make next year more accurate, not to prove this year was good.
• Supplement I: AI Literacy Learning Framework for Primary and Secondary Schools (p.1-4): sets out the three learning strands; the five-column table asks each row to name the strand it serves, and a row that cannot is itself a signal of “partly in place”.
• Supplement II: Guide to Using AI in Teaching in Primary and Secondary Schools (p.2-4): the teacher’s professional gatekeeping, compliance with the law, and safe learning; categories 2 and 4 carrying no names is that section’s boundary, applied to an internal review. (pp.2-2 to 2-5) The seven principles open with nurturing the whole person, with technology as an aid; the three judgements therefore turn on pupils’ learning and teachers’ professionalism, not on how much a tool was used.
• EDB Circular No. 156/2026 (Appendix 1, section 3): upholding professional gatekeeping; the review meeting is run by school colleagues, with no outside indicator standing in for a school’s own judgement.
• EDB Circular No. 107/2026 (paragraph 4): bringing digital education into the school development plan and/or the annual school plan, improving continuously through a self-evaluation cycle and reviewing the effectiveness of the work in good time; the five columns, three judgements and four points are that cycle’s record at school level. The Bureau prescribes no format for the review table, so these columns are a suggestion.

Learning strands: the fundamental operating principles of AI / human-AI communication and collaboration / reflection on the application of AI

Official documents page: https://www.edb.gov.hk/en/edu-system/primary-secondary/applicable-to-primary-secondary/it-in-edu/debp.html

This page is an independent Beets teaching resource, not an official EDB document. The name “annual AI review pack”, the five-column table, the six evidence categories and the three judgements are a suggestion from this resource, not an Education Bureau requirement; what the Bureau requires is continuous improvement through a self-evaluation cycle and reviewing the effectiveness of the work in good time, and it prescribes no format for the table. This page is for internal review only: it is not used for staff appraisal, it sets no ranking, and it compares the school with no other school. Anything paid for from the “Smart Learning with AI” grant must still submit its mid-term and final reports under that scheme’s guidelines; this page lists the interface only, does not replace either report, and gives no funding or financial compliance advice. This page collects no teacher’s or pupil’s personal data.

Feedback

If you have any comments or questions about this resource, or an idea for the next one, please tell us on the feedback page.