Key takeaways
- Choose a narrow teacher workflow before choosing a platform: planning, differentiation, feedback, resource creation, or administration.
- Publish an approved-use policy before staff upload student information or confidential school material.
- Pilot with a small, representative teacher group and measure time saved, output quality, corrections, confidence, and repeated use.
- Keep student-facing AI separate from teacher-preparation AI; the supervision and safeguarding requirements are different.
- Scale only when the pilot produces repeatable value and the school can support training, privacy, access, and review.
AI in schools works best when leaders treat it as a change to teaching practice, not a software launch. The useful first question is not Which AI should we buy? It is Which recurring job is taking time away from teaching, and what would a safe, better version of that workflow look like? A focused pilot makes the answer visible. A blanket rollout hides it.
Separate teacher AI from student AI
| Use | Typical examples | What leaders need to decide |
|---|---|---|
| Teacher preparation | Lesson drafts, worksheets, slides, rubrics, differentiation | Approved tools, curriculum fit, review expectations, student-data boundaries |
| Teacher administration | Summaries, communications, meeting notes, data analysis | Confidentiality, record retention, accuracy checks, who can access outputs |
| Student-facing support | Tutors, feedback, practice, brainstorming | Age requirements, consent, supervision, safeguarding, assessment rules |
| School operations | Policy drafting, procurement analysis, internal knowledge | Sensitive data, permissions, auditability, human approval |
Teacher-preparation tools can often be piloted with lower risk because teachers remain responsible for reviewing the output before it reaches students. Student-facing tools need a separate decision: students may interact directly with generated content, so age limits, supervision, assessment integrity, accessibility and safeguarding need explicit treatment.
What an AI policy for schools should cover
Approved uses
Name the tasks staff may use AI for and the tasks that still require a separate approval.
Data boundaries
State what student, family, staff, assessment, and confidential information must never enter an unapproved tool.
Human review
Make the teacher or school leader accountable for accuracy, bias, curriculum fit, accessibility, and tone.
Student disclosure
Explain when students may use AI, how they acknowledge it, and what counts as unacceptable assistance.
Tool approval
Define who reviews privacy terms, age rules, data processing, security, integrations, and retention.
Incident response
Give staff a clear route for reporting unsafe output, accidental data sharing, or inappropriate student use.
A six-step school AI pilot
- 1
Choose one expensive workflow
Pick a repeated job with a visible baseline, such as weekly lesson planning, differentiation, worksheet creation, or slide preparation.
- 2
Set the boundaries first
Agree what data can be used, who reviews outputs, which student-facing uses are excluded, and how concerns are reported.
- 3
Recruit a representative group
Include confident and cautious teachers, different subjects, different year levels, and at least one curriculum or school leader.
- 4
Train on real work
Use actual planning tasks rather than generic prompt demonstrations. Show teachers how to review, correct, adapt, and reject AI output.
- 5
Measure value and failure
Track time saved, repeated weekly use, corrections required, curriculum fit, teacher confidence, and any privacy or safeguarding issues.
- 6
Scale, change, or stop
Expand only the workflows that teachers repeat and leaders can govern. A low-use pilot is evidence, not a reason to force adoption.
How to evaluate an AI platform for schools
Ask vendors to demonstrate your school's real workflow—not a polished generic prompt. Bring one curriculum objective, an existing resource, a differentiation need, and your required output format. Then check:
- Curriculum fit: Does the platform understand the framework your teachers actually use?
- Editability: Can teachers correct and adapt the output without starting again?
- Connected workflow: Can a lesson become a worksheet, slide deck, image, rubric or follow-up resource?
- Administration: Can leaders manage access, understand adoption and remove users cleanly?
- Privacy and security: Are data use, retention, subprocessors, training practices and contractual commitments clear?
- Accessibility: Can teachers produce materials that remain usable for varied learners?
- Exit: Can the school export its work and leave without losing essential teaching resources?
Pilot one connected planning workflow
Kuraplan gives schools a focused workspace for curriculum-aligned lessons, units, worksheets, slides, images, and teacher collaboration.
See Kuraplan for schoolsFrequently asked questions
Good starting points are teacher-controlled, repeated tasks such as drafting lessons, adapting materials, creating worksheets, preparing slides, writing rubrics, and summarising non-sensitive information. Teachers should review every output before use.
That is a school or district decision. Leaders should review the specific educator product, privacy terms, data agreement, age rules, local policy and intended use rather than approving a brand name in the abstract.
Do not enter personally identifiable student information, protected records, safeguarding information, confidential staff material, unpublished assessments, credentials, or anything prohibited by school policy.
Long enough to cover repeated real work. Four to six teaching weeks is usually more informative than a single training day because it reveals whether teachers return to the workflow after the novelty wears off.
Measure time saved, repeated use, output quality, corrections required, teacher confidence, curriculum fit, and any privacy, safeguarding or accessibility problems. Avoid counting logins as impact.