Pilot Overview

The pilot phase ran the school's new tool approval workflow on three live tools: Google Workspace for Education (Gemini AI), Canva for Education (AI features), and Edmodo. This generated the first real evidence of the governance framework operating under real conditions and established the KPI baseline for Year 1 reporting.

3
Tools Through Approval Workflow
2
Approved (with conditions)
1
Suspended Pending DPA
4
Policy Refinements Triggered
0
P1 Incidents During Pilot
Tool Approval — Workflow Outcomes
Integrated Platform · AI Features
Google Workspace for Education Plus — Gemini AI
✓ Approved with Conditions

Review Findings

  • Google confirmed student data in Education Plus tier is not used to train Gemini models — DPA clause verified.
  • Gemini features for students under 14 require admin restriction in Workspace console — completed Week 1 (emergency action).
  • Smart Compose and Smart Reply: approved for all ages — no student data processed externally.
  • Gemini in Google Docs/Slides: approved for Grades 8+ with teacher supervision; restricted for Grades 1–7.
  • Gemini in Meet (transcription): approved for staff only.

Approval Conditions

Admin console reviewed and restricted before each new academic year or Google product update.
Parental notification letter issued and archived before student-facing Gemini features are enabled.
Grade 8+ Gemini use requires teacher to set task context — open prompting not permitted in class.
Annual equity audit to include review of Gemini outputs for EAL student bias.

Data Processing Agreement

Google Workspace for Education DPA signed. Data residency: US (Google Cloud). Ley 1581 cross-border transfer clause included. ✓

Creative Platform · AI Features
Canva for Education — AI Features (Magic Media, Magic Write)
✓ Approved with Conditions

Review Findings

  • Canva for Education tier confirmed: AI-generated content is not stored or used for model training.
  • Magic Media (AI image generation): suitable for Grades 4+ with teacher oversight. Requires content guardrail review.
  • Magic Write (AI text): suitable for Grades 6+ as a drafting aid; not for submission as final work without disclosure.
  • DPA reviewed and signed — Colombian data transfer clause added.

Approval Conditions

Magic Media (image generation) restricted to teacher-assigned projects only — not open student access.
AI-generated images in student work must be labelled "created with AI" before submission.
Magic Write disabled for Grades 1–5. Enabled for Grades 6–12 with teacher framing.
Annual review of Canva AI feature set as platform evolves.
LMS · AI Features
Edmodo — AI Content Recommendations & Analytics
⏸ Suspended — DPA Pending

Review Findings

  • Edmodo's updated privacy policy (Jan 2026 revision) contains an ambiguous clause permitting "aggregate and de-identified data to improve AI recommendation systems."
  • DPA template submitted to Edmodo — awaiting response as of report date.
  • No evidence of active breach; precautionary suspension applied.
  • Affected classes (3 teachers, ~90 students) migrated to Google Classroom.

Next Steps

Await Edmodo DPA response (deadline: March 31, 2026).
If DPA satisfactory: conditional approval with data minimisation settings required.
If DPA rejected or no response: permanently remove from approved list.
Governance Panel to review at [Quarter Year] meeting.
KPI Baseline — Day 60 (Post-Program Launch)

These are the baseline values established at Day 60 of the program launch (February 2026). They form the reference point for all future governance reporting.

[X]%
Staff training completion
Target: ≥80% ✓
3
Tools through approval workflow
Target: ≥2 ✓
[N]%
Leadership team trained
Target: [target]
[N]
Incidents logged (all P2/P3)
[status]
[N]
P1 critical incidents
Target: [target]
[N]%
Parent session attendance
Target: [target]
[X]%
Parents: "trust school on AI"
Baseline established
[R.R]/5
Governance readiness score
Was [B.B] — change [±X.X]
Equity Audit — Student-Facing AI Tools

Each approved student-facing tool was assessed across 5 equity dimensions. No critical failures identified. Three areas flagged for monitoring.

Equity Dimension Finding Status
EAL / ESL StudentsGemini AI outputs reviewed for 15 EAL student samples. No systematic bias in feedback quality detected. Spanish-language prompts handled accurately.Pass
Students with Learning DifferencesAI text-to-speech and reading support features in Schoology and Google Workspace reviewed. Accessibility features intact and functional.Pass
Device / Access ParityGrades 1–3 (shared devices): AI features configured for teacher-led use only — no unsupervised student access. Grades 4+ (1:1): full access per policy.Pass
Gender Bias in AI FeedbackSample review of AI-generated comments across 40 student writing submissions (mixed gender). No systemic tone or language differences detected. Monitoring recommended.Monitor
Socioeconomic Access (Home Use)3 families identified without reliable home internet. AI homework tools restricted to school use for those students. Support pathway activated.Monitor
Policy Refinements (v[X.X] → v[X.X])

Four refinements were made to the Living AI Policy based on pilot findings. Policy updated to v[X.X] on 13 February 2026.