| Apollo | A typical RTO platform | |
|---|---|---|
| What its AI reads | The submitted document: Word, PDF, portfolios, scans | Answers typed into its own question fields |
| Who makes the judgment | Your assessor, on every section, before anything is final | Varies. Often a suggested outcome an assessor accepts in bulk |
| What the feedback says | Quotes the work, names the guide’s requirement, points to the sample answer, says what to change | Varies. Commonly an outcome against the criterion, with the wording left to the trainer |
| Where student work sits | Nowhere. Held in memory, cleared on close | On their servers, as the system of record |
| Who pays for the AI | You do, directly, at cost. 9.2c an assessment | Bundled into a per-learner price |
| Integrity checking | Across the class, on your machine, nothing uploaded | Generally not offered |
| To adopt it | Install it. Nothing to migrate, because nothing goes in | Migration, re-authoring assessments, cutover |
| What a trainer learns | One screen: drop the file, review the marks, sign off | A whole platform |
| To stop using it | Uninstall. There is nothing of yours to extract | Your records live in their schema |
Why we call it an assessment judgment assistant.
Three words, each doing work: it reads assessment evidence, prepares the judgment, and assists rather than decides. “AI marking” would be easier to sell. It would also describe a different product, one that puts your RTO on the wrong side of an auditor’s question.
What Apollo is not.
It is a marking instrument and deliberately nothing more.
Not a student management system
No enrolments, no certificates, no AVETMISS reporting. Apollo works alongside whatever you already run.
Not a plagiarism database
The integrity check compares the class in front of it and the sample answer. It cannot see the web or previous cohorts.
Not the decision-maker
Apollo drafts; a qualified person judges, confirms and signs.
Not a validation service
It gives your validators something explicit to review. It does not do the validating.
How we measured, and what we haven’t.
Every figure on this site comes from a real run. Here is the method, including its limits.
A trainer at the college where Apollo was built put the same pile at 6.5 hours by hand. Her two hours included admin she would have done anyway. The two hours were timed. The 6.5 is her estimate.
Forty students, two tasks, three units. That is 65 hours of marking by hand against 20 with Apollo, before resubmissions. Your own minutes will differ.
Measured on a real project task marked against a full unit context.
What has not been proven: marking quality has not been independently validated, and the timing is one trainer, one batch, one assessment type. Real evidence, not yet a cohort study.