AI that assists. Teachers who decide.
Every grade the AI suggests is reviewed, overrideable, and only final when you lock it.
- You approve every grade
- Student data protections
- Plain-language AI policy
The Grade Harbor AI promise
The AI prepares a structured draft so teachers can spend their review time on judgment, context, and final decisions.
- AI suggestions are drafts, never final grades.
- Teachers review, edit, override, and lock every grade.
- No grade passes to students without teacher approval.
- We do not use Student Data to train generalized AI models without explicit written authorization from the applicable school, district, or authorized educational institution.
- Provider, model, and prompt metadata are retained where auditability matters.
How grading works
How grading and teacher review work
The grading pipeline keeps the rubric, AI reasoning, teacher edits, and final lock decision in one visible review flow.
- 01
You define the rubric
The AI grades to your criteria, point values, and anchor descriptions instead of applying a generic standard.
- 02
The essay is evaluated against that rubric
Prompt-backed grading uses the rubric snapshot and student writing to produce criterion-level draft scores.
- 03
You see the AI's reasoning
Every score includes written reasoning and confidence guidance so teachers can inspect the evidence.
- 04
You edit or override
Teachers can revise any score or feedback. The AI draft is a starting point, not the answer.
- 05
You lock the final grade
Only the teacher-approved, locked version becomes the final grade and feedback record.
AI draft
Teacher review workspace
Criterion draft
Needs closer reviewThe draft score includes AI reasoning, confidence guidance, and editable teacher controls before anything is locked.
Boundaries
What the AI can and cannot do
The page is explicit about both capabilities and limits so teachers and compliance reviewers can evaluate the workflow without guesswork.
What the AI can do
- Score essays against a rubric criterion-by-criterion with written reasoning
- Generate written feedback grounded in the specific essay text
- Flag scores that need closer review when confidence is low
- Identify class-level writing patterns and skill gaps
- Suggest groupings and instructional priorities based on skill profile data
What the AI cannot do (and does not try to)
- Make a grade final without teacher review - the system prevents this
- Assess tone, intent, or context that requires knowing the student
- Evaluate factual accuracy outside writing craft, such as whether a citation is true
- Replace the teacher's professional judgment about a student's situation
- Communicate with students - only the teacher can share feedback
Student data
What happens to student essays
These statements mirror the current legal and architecture docs without making new legal commitments.
Student data handling
- Essays may be processed by trusted AI service providers for grading and feedback.
- Essays are stored securely in our system tied to your account and protected by access controls.
- We do not use Student Data to train generalized AI models without explicit written authorization from the applicable school, district, or authorized educational institution.
- Teachers and schools can request deletion through the product or support process, subject to applicable school agreements and automated retention workflows.
- Where applicable, we operate as a school service provider under FERPA's school official exception.
Legal and policy references
For full compliance details, review the current policy pages. This AI Safety page is a plain-language summary of current product behavior and policy references.
Policy details
Where AI provider details live
This page explains the teacher workflow. Current provider and model details stay in the AI Use Policy so compliance references do not drift.
Provider and model policy
- AI service provider details are maintained in the AI Use Policy so procurement and legal reviewers have one current source.
- The system stores provider, model, prompt version, and prompt hash metadata where auditability matters, including grades.
- We do not fine-tune generalized AI models on Student Data without explicit written authorization from the applicable school, district, or authorized educational institution.
- If AI workflows, providers, or model configurations materially change, we provide notice as described in the AI Use Policy and applicable agreements.
We do not make specific claims about model accuracy that we cannot verify. Rubric-based AI grading is assistive; it is not infallible.
Read the AI Use PolicyWhen to slow down and review carefully
Needs closer review is the teacher-facing label for low confidence. It means the AI found limited, conflicting, or unusual evidence and wants you to inspect that score before finalizing.
High-confidence scores can often be reviewed faster, but the teacher still makes the final call.
Review carefully when you see
- Any score marked "Needs closer review"
- Scores for unconventional writing styles
- Essays from multilingual learners
- Very short or incomplete submissions
- Assignments requiring outside factual accuracy
FAQ
Questions and concerns
Short answers for teachers, school leaders, and reviewers evaluating whether AI belongs in a grading workflow.