The PIE Cycle Meets AI Dashboards: Evidence for School Planning, Implementation and Review
Every school runs Plan–Implement–Evaluate, but evaluation often relies on impressions. See how PIE KPIs link directly to platform data — AI usage, objective mastery, teacher training records, SDAS results — so planning has a baseline, implementation has tracking, evaluation has evidence, and grant interim and final reports export from the same cycle.
At a June panel meeting, the maths head turns to last September's goal: "Improve pupils' self-directed learning." The principal asks, "What figure?" Silence for five seconds. The following week the same school opens the Edor.ai PIE dashboard: P4 tutor-task completion rose from 38% to 61%, and the median SDAS standard score gained 0.27. "Review finally has material — we are not guessing."
PIE is not a new framework. It fails when evaluation has no data. Link KPIs directly to platform records and planning gets a baseline, implementation gets tracking, evaluation gets evidence — and the grant's interim and final reports become by-products of the cycle instead of paperwork invented in the reporting season.
Three breaks in school PIE
Hong Kong schools write plans every year and hit three predictable breaks. Planning sets abstract goals — "embed AI in teaching" with no baseline, so there is nothing to measure from. Implementation lacks tracking — in March you discover a panel never started, too late to recover. Evaluation runs on impressions — "felt positive" and "good response" at the year-end meeting, with no answer for the management committee's "what changed, exactly?"
An AI dashboard is not another system to maintain. It threads the three stages with one dataset: the baseline you read in September is the endpoint you compare in June.
Writing KPIs the system can compute
Vague KPIs cannot drive a dashboard. Rewrite goals as computable sentences with a named data source:
- "Teachers use AI more for planning" → "≥ 70% of teachers use AI planning tools at least twice a month" (source: usage records by role / module)
- "Pupil self-directed learning" → "≥ 60% of P4–P6 pupils complete at least one tutor task weekly" (source: tutor task completion)
- "Raise maths results" → "P4 Maths standard score improves ≥ 0.3 from term 1 to term 2" (source: SDAS)
- "Teacher professional development" → "Three trainings, three demo lessons and three sharing sessions completed, each with date, headcount and hours logged" (source: training records)
- "Cross-subject use" → "AI resources tagged as teaching exemplars used in ≥ 3 subjects × 2 grades" (source: library + usage records)
Once KPIs read like this, the dashboard can update attainment daily and a panel head sees a lagging indicator in October, not June.
Year one: few and computable
In year one, choose five KPIs, all computable, with at least two on teacher behaviour (usage, training) and two on pupil outcomes (mastery or SDAS). Too many targets and the dashboard becomes decoration nobody opens.
What each stage looks at
Plan (September)
The dashboard shows baseline: last year's AI usage by panel, objective mastery by grade, SDAS starting points. The principal and AI lead draft this year's KPIs from figures, not from thin air — if last year's P4 tutor completion sat at 38%, a target of 55% is more honest than "significant improvement" and easier to defend in June.
Implement (October–May)
The dashboard switches to tracking: live attainment per KPI, alerts when a panel sits below threshold for two consecutive months, a monthly automatic summary attachable to school meetings. Permissions tier — principal sees the school, panel heads their subject, teachers only their own indicators. Data is for discovery and dialogue, not league tables.
Evaluate (June)
The dashboard produces an evidence pack: KPI attainment table, before/after charts for key indicators, teacher and pupil feedback summaries. AI can draft evaluation narrative, but the panel head must edit — numbers explain what happened; professional judgement explains why and what comes next.
Put fifteen minutes of "data time" on the monthly panel agenda. Without rhythm, nobody opens the dashboard after month three.
Mapping PIE to the grant report
The EDB "智" 啟學教 funding (up to HK$500,000 per school, programme deadline 31 August 2028) requires an interim report (October 2027) and a final report (October 2028). The PIE dashboard can export matching data fields:
- Scope of AI use (subjects, grades) ← usage by subject × grade
- Teaching exemplars ← library items tagged "exemplar"
- Teacher professional development ← training, demo and sharing records (date, headcount, hours)
- Pupil learning outcomes ← objective mastery, SDAS before/after
- Sustainability ← number of school resources, shared teacher templates
Export Markdown or CSV in the report structure, then the school adds narrative and reflection — see how to use the EDB HK$500,000 AI grant. The point stands: if you do not run PIE during the year, reporting season becomes a record-chasing ordeal.
Permissions, culture and common traps
A dashboard only works if someone looks. Three lines: data supports panel conversation, not public ranking of teachers; a lagging indicator triggers "what support is needed", not blame; the principal and AI lead read the school summary monthly to model data in decisions.
Common traps: KPIs set so high the dashboard stays red all year and nobody trusts it; tracking AI usage alone without learning outcomes, so the grant's "impact" section has nothing to say; evaluation meetings that still trade anecdotes without projecting the dashboard — which makes the work pointless.
Common misreadings
-
The claim: A dashboard equals automated school management. What actually happens: The dashboard computes KPIs and exports data, but goal-setting, interpretation and action stay human. Without fifteen minutes of data time each month, it becomes a static screenshot.
-
The claim: The grant report can be 100% automatic, so June meetings disappear. What actually happens: Data fields export; narrative, reflection and management-committee approval remain the school's work. The dashboard saves record-hunting time, not professional judgement.
-
The claim: More KPIs show the school is serious. What actually happens: Five computable KPIs in year one is enough. Too many indicators scatter panel attention, and some lack reliable sources — which sends you back to manual forms.
Three things to try on Monday
- Open any panel plan from last year and rewrite one vague goal as "who, what proportion, doing what, data from where" — for example "70% of teachers use AI planning ≥ 2 times/month / usage records".
- With your AI lead, draft five year-one KPIs and confirm each names a platform module source (usage, mastery, SDAS, training, library).
- Add "15 minutes data time" to the next panel agenda. Even before full platform integration, log one baseline figure in a spreadsheet — you need a September starting point to have a June comparison.
Questions that keep coming up
No. Plan–Implement–Evaluate is a framework you can apply to the annual plan, a panel plan or an AI rollout project. What matters is whether KPIs can be computed automatically from the platform rather than marked "achieved" from memory at a year-end meeting.
For example, a KPI stating "≥ 70% of teachers use AI lesson planning at least twice a month" is calculated directly from usage records by role and module. "P4 Maths standard score improves ≥ 0.3 from term 1 to term 2" is computed automatically by the SDAS module — no panel secretary chasing individuals.
The data sections can export in the report structure — subject and grade coverage, training records, teaching exemplars, mastery and SDAS before/after comparisons, and more. Narrative, reflection and management-committee approval still belong to the school. The dashboard organises evidence; it does not replace professional judgement.
Related notes
SDAS Student Data Analysis: Eight School Decisions Beyond the Mark Sheet
A mark sheet tells you who scored what; a Student Data Analysis System answers why, what the trend is and what to do next. Eight questions show how standard scores, percentiles, progress rates, multi-year tracking and teacher analysis are used — and how AI natural-language queries let every teacher ask.
Keep reading
Reclaiming 8–10 Hours a Week: AI Admin Agents for Notices, Minutes and Parent Letters
Hong Kong teachers spend about 8–10 hours a week on paperwork. Six eAdmin scenarios show how AI drafts eNotice, eCircular, parent letters, meeting minutes, event proposals and iPortfolio comments — and why grounding in the school knowledge base and matching the school's voice decide whether a first draft is usable.
Keep reading
AI Tools for Teachers in 2026: Where Nine Products Actually Sit, Who Their Terms Exclude, and What They Cost a Whole School
Gemini, Gamma, Edpuzzle, Khanmigo, Perplexity, OpenAI Codex, Claude Code, Cursor and Base44 are not competing on the same track. Three lines separate them: whether the product treats you as a consumer, a developer or an institution; whether its terms let students in; and whether it bills per person or per school. Official September 2026 figures, plus the arithmetic for 100 staff.
Keep reading