AI workflow
Six stages, and a teacher decision in the middle of every one
The workflow is deliberately transparent: you can always see what the AI proposed, what evidence it used, and who approved the result.
Stage by stage
From classroom evidence to an approved lesson
Each stage states who acts and exactly what the AI is permitted to do.
- 1
Teacher
Capture evidence
Assessment results, performance tasks, and classroom observations are recorded against competencies.
What the AI may do
Structures and tags incoming evidence. Makes no judgement about the learner.
- 2
Platform
Update learner models
Mastery, misconceptions, pace, and confidence are recalculated for each learner and competency.
What the AI may do
Explains what changed and which evidence caused the change.
- 3
AI Coach
Propose the next lesson
A draft plan is generated for the competencies your class has not yet mastered.
What the AI may do
Drafts only. The plan stays in draft until a teacher reviews it.
- 4
Teacher
Teacher review
The teacher edits, rewrites, or rejects any part of the draft and approves the final version.
What the AI may do
Records the decision and the reasoning for future suggestions.
- 5
Teacher
Teach and assess
The approved lesson is delivered with differentiation and interventions already prepared.
What the AI may do
Prepares aligned assessment items for the competencies taught.
- 6
Teacher & platform
Reflect and improve
Outcomes are compared with expectations, and the next cycle starts from stronger evidence.
What the AI may do
Reports what worked, what did not, and how confident it is in each conclusion.
The teaching loop
Observe, analyze, plan, teach, assess, reflect, improve
Every cycle feeds the next, so lessons stop being generic the moment your learners produce evidence.
- 1
Observe
Capture classroom evidence, assessment results, and teacher notes.
- 2
Analyze
Update mastery, misconceptions, pace, and confidence per learner.
- 3
Plan
Draft adaptive lessons tied to official competencies and real evidence.
- 4
Teach
Deliver with differentiation, interventions, and enrichment prepared.
- 5
Assess
Measure the competency you targeted, not a generic quiz.
- 6
Reflect
Record what worked; the platform learns your teaching preferences.
- 7
Improve
Every next lesson starts smarter than the last one.
Guardrails
The rules the AI operates under
These constraints are product requirements, not settings a school has to remember to switch on.
Evidence-based
Every recommendation cites the learner data it came from — no unexplained output.
Confidence-scored
Each suggestion carries a confidence level and its stated limitations.
Editable
Nothing is locked. Teachers rewrite, trim, or discard any AI draft.
Teacher-approved
No instructional recommendation is ever applied automatically.
Security is part of the instructional design
- Row-Level Security on every business table, enforced server-side.
- Teachers reach only their assigned classes; students only their own records.
- Privileged and administrative actions are written to an audit log.
- AI reads only the data the requesting user is authorised to see.