SAGE — Structured AI-Guided Education Framework

Why SAGE

Permission is only the beginning

A permission setting tells students whether they may use AI. SAGE carries that decision all the way through: SAGE-AP turns institutional permission into clear assessment conditions, the six-step pedagogy teaches students how to use AI responsibly, and a supervised Defend step confirms they understood what they produced. Validated with more than 1,500 students across 30+ studies in Australia, China, and the United Kingdom.

📋

Teaches students how, not just whether

An AI policy says "you may use AI at Level 3." That tells students nothing about how to use it well. SAGE gives them a repeatable process: check the AI's work against real sources, fix what is wrong, and show their reasoning. The thinking has to be theirs.

🔬

Reveals what students actually know

When a student cannot explain what the AI got wrong, cannot point to a source that contradicts the AI, or cannot defend their own changes — that tells you something important about their learning. SAGE makes these gaps visible, not hidden.

🛡️

Closes the verification gap

In a structured audit, only 12% of unsupervised submissions produced a genuinely traceable evidence trail — even with process logs and AI interaction records. That is why SAGE includes a Defend step: a short, supervised checkpoint where each student shows they can explain and justify their own work.

SAGE alongside other frameworks

Within the SAGE programme, SAGE-AP provides the permission-to-pedagogy alignment layer: it translates institutional permission settings into clear conditions for responsible AI-supported assessment. SAGE Pedagogy then structures how students generate, evaluate, refine, use AI as critic, reflect, and defend their work. Together, SAGE-AP and SAGE Pedagogy connect permission, practice, and assurance.

The AI Assessment Scale (AIAS) remains a compatible external permission scale. Institutions may use AIAS to express permitted levels of AI use, while SAGE-AP aligns those settings with the SAGE process and student-facing assessment conditions.

Read the implementation guide

Dimension	Permission layer	SAGE
Answers	How much AI?	How to use AI?
Unit	Permission level	Process cycle
Student action	Follow the rule	Apply the process
Assurance	Not addressed	Defend step
Validation	Varies by framework	1,500+ students, 30+ studies

How it works

Six steps, one assessment

SAGE is not an add-on to your assessment. It is the assessment. Students work through six structured steps — five using AI openly, and one final checkpoint under supervision.

The SAGE Cycle

Open

AI-integrated tasks

1

Generate

→

2

Evaluate

→

3

Refine

→

4

AI Critic

→

5

Reflect

Assurance

Supervised verification

6

Defend

Steps 1–5 build capability through structured AI-supported learning. Step 6 assures individual attainment through supervised verification.

Step 1

Generate

Students use an AI tool to produce an initial response to the task. In early weeks you provide the prompts. Later, students write their own — but must first show they understand the problem before engaging the AI.

Step 2

Evaluate

Students compare the AI output against real sources — industry standards, peer-reviewed research, clinical guidelines, or whatever applies to your discipline. They identify what is correct, what is missing, and what is wrong.

Step 3

Refine

Students fix the AI output based on their evaluation. Every change is documented: what they accepted, what they modified, what they rejected — and why. The result is work the student can genuinely call their own.

Step 4

AI Critic

Students flip the relationship. They tell the AI to act as an expert reviewer and then judge whether the AI's critique is valid or whether the AI itself is wrong. This is where students develop real authority over the tool.

Step 5

Reflect

Students write a structured reflection: what the AI got right, where it failed, what domain errors they found, and what this tells them about when AI can and cannot be trusted in their discipline.

Step 6

Defend

At one controlled moment — a short viva, a live demo, a timed exercise — each student proves under supervision that they can explain and justify their work. This closes the gap between "submitted good work" and "actually understood it."