Marking that takes
a night, not weeks.
Two OCR engines read every script and two AI graders mark it, with a third arbitrating when they disagree. Your Subject Matter Expert signs off every mark. Built for CAPS, DBE and IEB.
- CAPS curriculum · DBE and IEB exams
- No single AI reads or marks a paper
- Audit-ready evidence per mark
- POPIA-aligned
- Schools + tertiary
Product walkthrough
How does it work?
Step 01
Marking guide setup by a Subject Matter Expert
Your Subject Matter Expert verifies the marking guide once per exam. Criteria, mark codes, tick anchors, alternatives, and rubric levels are all editable. This SME-verified setup becomes the single source of truth. The AI defers to it.
Your call, not ours
Your marking guide is the authority
Your Subject Matter Expert sets what earns a mark, including the alternative wordings a human marker would accept. Nothing is marked until they have signed the guide off.
Step 02
Dual OCR runs
Two independent OCR engines transcribe every script before any marking begins. Typed and handwritten papers are both supported. Where the two transcriptions disagree, the question is flagged for your Subject Matter Expert to read the original, rather than resolved by a guess.
What you do
Upload the batch. That is your part.
Drop in one PDF or a hundred. A single script is just a batch of one. Nothing can be marked against a guide nobody has checked, so the upload sits behind the verified setup.
Step 03
AI grading
Every paper is independently graded by two AI examiners. Where they disagree above the threshold you set, a third AI arbitrator adjudicates the specific question. No question is marked by one model alone, which reduces single-model bias.
AI Grader 1
5/ 10AI Grader 2
8/ 10Arbitrator
8/ 10Resolution
No single AI decides
Marked twice, then arbitrated
Two AI examiners mark every question independently. Where they disagree, a third adjudicates against your marking guide rather than splitting the difference. No mark on your papers is decided by one model on its own.
Step 04
Moderation by a Subject Matter Expert
Papers land in a moderation dashboard sorted by confidence. Your SME reviews flagged questions, accepts or corrects marks, adds comments, and places ticks or Xs directly on the script. Every change is logged to the audit trail.
What comes back
Red is the AI, blue is your SME
Marks land on the script itself, not in a spreadsheet of totals. Your Subject Matter Expert marks in blue over the AI in red, so who awarded what is never in question, and the running total moves as they work.
Step 05
Analysis and handover
Download graded scripts, cover pages, examiner reports, per-question statistics, reliability metrics, and a full audit JSON, all in one ZIP. Student data is purged 72 hours after you download the handover bundle, unless your institution has set a longer window. You own everything that leaves the platform.
More than marks
Where the cohort actually landed
Every batch comes back with the class plotted against NSC grade boundaries, so a set of marks becomes something a head of department can act on.
Yours to keep
You own everything that leaves
Marks with a per-question breakdown, the annotated scripts, an examiner report, and a full audit trail behind every mark. It is your institution's data and it leaves in formats you can open without us.
Why ExamMark
Built for South African institutions, not adapted to them.
Built for South African assessment
CAPS, DBE, and IEB grading conventions are wired into the system, including the SA mark-code set and SA rubric formats. ExamMark was built around them rather than adapted to them afterwards.
Dual-grader arbitration
Two independent AI graders assess every paper. A third arbitrator resolves disagreements above a threshold you set. It is the same principle as the moderation step in formal SA examinations: no mark rests on one judgement.
Human authority
Human in the loop
The AI suggests. The Subject Matter Expert decides. No mark is final until a Subject Matter Expert signs off in the moderation flow. Override any mark, any time, with no friction.
When the AI graders disagree
An arbitrator adjudicates, using your marking guide.
AI Grader 1
5/ 10AI Grader 2
8/ 10Arbitrator
8/ 10Resolution
Example reason: Grader 1 did not credit a phrasing the marking guide lists as an accepted alternative. The arbitrator checked the guide and upheld the higher mark.
The Subject Matter Expert reviews and signs off the final mark.
How we work with you
One managed platform, tailored to your institution. Pricing is scoped per institution. Book a conversation and we will talk through what that looks like for yours.
Managed Platform
We run it, tailored to your institution.
We host, tune, and operate the platform on your behalf. Marking guides, moderation flow, and reporting are shaped around how your institution already works. Your team signs off the marks; we handle everything else.
Best for institutions that want the output without the operational lift.