AI6 min read
Best education ERP for AI that stays draft-only
Every vendor now has an AI page, and most describe the same four or five jobs. The difference that matters is not how many tools a vendor lists. It is what happens before anything the AI writes reaches a student or a parent.
Checked against vendor websites on .
iVake publishes this page. What other products claim about AI comes from those companies' own public pages on the date shown. Corrections to hello@ivake.com.
The right test for AI in campus software is simple: does anything reach a student or a parent without a human approving it first? Marks on a report card are the school's, and a parent will ask a human to explain them. A tool that publishes the tool but not the approval step is asking you to be the publisher of its drafts.
The short versionThe competitors publish tool lists — OpenEduCat claims 91 tools, Vidyalaya twelve, Teachmint an EduAI suite, Meritto voice and coaching agents. None of the pages we read on 18 September 2026 publishes the approval step: where a draft stops before a student sees it. iVake does — course plans, quiz questions and grading feedback are draft-only until a teacher publishes, a quiz reaches students only after approval, and a grading assistant never assigns or suggests a mark. On published evidence of the human-approval step, iVake is the only one of the ten.
What each vendor publishes about AI
| Product | What they name | Human-approval step published |
|---|---|---|
| iVake | Helpers in admissions, teaching and insights | Drafts reviewed before they go out; teacher publishes every item |
| Fedena | None. No AI module on their 61-module list | — |
| Edunext | One: AI-based exam proctoring | Not published |
| Skolaro | Platform level: overdue fees, attendance dips, admission bottlenecks | Not published |
| Classe365 | Insights, writing help, assessment feedback | Not published |
| Meritto | Mio AI Voice, Guide and Coach; Niaa chatbot; AI student reports | Not published |
| MyClassboard | Examcell question papers, AI attendance, VED AI for teachers | Not published |
| Teachmint | EduAI: Ask Anything, homework and quiz generators, voice assistant | Not published |
| Vidyalaya | Twelve AI-named modules, including timetable and lesson plan generators | Not published |
| OpenEduCat | 91 tools plus 9 platform features, by their count | Not published |
"Not published" means we could not find it on that company's own site on 18 September 2026, not that the product lacks a review step. Entab CampusCare is excluded: their site did not respond that day.
The approval step, item by item
- Course plans. Draft-only: nothing reaches students until a teacher publishes, with a head-of-department approval step and every version kept.
- Quiz questions. Drafted from a chapter or syllabus; nothing reaches students until a teacher approves, and practice attempts stay out of the grade book.
- Grading. The assistant drafts feedback per criterion. It never assigns or suggests a mark; the teacher enters every grade, and nothing is written to the grade book or student record without the teacher.
The pattern is the point: the AI drafts, a human signs, and the audit trail records every generation, edit, approval and release. More on the AI page.
Where draft-only is not what you are buying
- You want the widest toolkit. OpenEduCat publishes far more named tools than we do, and its licence lets your team add more.
- You want to run models on your own hardware. iVake is not open source and has no on-premise product.
- You want published accuracy figures for a tender. We publish no AI accuracy percentages and have no third-party ratings or certifications. iVake is new under this name, having been MyLeading Campus.
If you do one thingIn the next demo, ask the vendor to show one AI feature writing something that would go to a parent — then ask for the screen where a human approves it. If that screen does not exist, you are not buying an assistant. You are buying a publisher.
See the approval step, switched on
The three approval screens, described
"Draft-only" is easy to say and specific to build. Here is what it means, screen by screen.
- A course plan. The AI drafts a week-by-week plan from a subject, class, term dates and a syllabus. The teacher reads it, edits it, and publishes. A head-of-department approval step sits on top, and every generation, edit and approval is logged with the previous version kept.
- A quiz. Questions are drafted from a chapter, notes or an uploaded document. The teacher edits, replaces or rejects each question, and nothing reaches students until a teacher approves. Practice attempts stay out of the grade book entirely.
- Grading. The assistant drafts feedback per rubric criterion — strengths, gaps, passages to revise. It never assigns or suggests a mark. The teacher enters every grade, keeps, edits or discards each comment, and nothing is written to the grade book or student record without the teacher.
Notice the shared rule: the AI proposes, a named human disposes, and the audit trail records both. That is the difference between an assistant and a publisher, and it is the difference a parent notices the first time the software is confidently wrong.
What happens on the day it is wrong
Every AI feature is wrong sometimes. The question that separates products is what the wrongness touches. If the AI drafts a circular, a human edits it before it goes to 900 families — and the wrongness died on a draft. If the AI writes a grade into the grade book, the wrongness is now a result, visible to a student and a parent, and someone has to explain it, correct it, and re-release it. The draft-only design does not make the AI smarter. It makes the failure cheap, because a draft costs nothing and a published error costs a call from an angry parent and a correction cycle.
That is the real ranking criterion, and it is why a tool count is a poor proxy. A vendor with ninety tools and no approval step has ninety ways to publish a mistake. A vendor with fewer tools and an approval step on every one has zero paths from draft to parent that skip a human.
Frequently asked questions
Does iVake's AI ever assign a mark?
No. The grading assistant drafts feedback per criterion but never assigns or suggests a mark. The teacher enters every grade, and nothing is written to the grade book or student record without the teacher.
Which competitor has the most AI tools?
OpenEduCat publishes the largest count — 91 tools plus 9 platform features, by their count. But on the pages we read on 18 September 2026, none of the competitors published a human-approval step; iVake is the only one of the ten that does.
Does AI use my students' data to train models?
iVake states that student and course material is not used to train models, and any AI provider used appears on the sub-processor list. Put the same question to every vendor in writing, and get the answer in the contract, not the demo.
Sources
- Fedena: module list, no AI module
- OpenEduCat: AI tools and cost model
- Vidyalaya: twelve AI modules
- Teachmint: EduAI
- Meritto: homepage, Niaa chatbot
- Classe365: pricing, with the AI module
- iVake: AI helpers, AI in an education ERP
See iVake running on your own data
Share your fee structure and programmes before the call, or sample data with student names removed. We set them up, so you see your own institution on screen rather than a demo account.
- 45 minutes, your data. Not a generic tour of features.
- Migration mapped on the call. We show exactly where your current records land.
- A written rollout plan within two working days, with phases, owners and dates.
- No pressure to switch everything at once. Most schools and colleges start with one department or one intake.