The AI marks
in pencil.
Your teachers
sign in ink.

It can suggest one, which an examiner then accepts or changes — it never marks unattended. A teacher can ask for suggestions on one script or on the whole class at once. It copes with pages photographed out of order and with an answer the student numbered wrongly, and it credits any valid method, not only the one in the model answer. Published research on automated grading puts the share any system can safely grade without a person at 11-20%, so this is built as a first draft under every question rather than a replacement for one. It only runs when an institution has switched it on for a specific exam, and nothing about a student's behaviour, camera or microphone is ever part of it. When it is on, photographs of the answer sheets and the questions are sent to an AI provider, OpenAI or Anthropic, which both state they do not train on this data and delete it within 30 days; nothing else about the student is sent. Beside every suggestion the examiner sees what the AI read on the page, and the page it read it from, before deciding the mark.

Yes. On the results page one button queues every script that has no suggestions yet, and a progress bar shows how far it has got. Scripts are read one at a time in the background, so the page can be closed, and a script that could not be read is offered again. Every suggestion still waits for a teacher to accept or change it - there is no button that turns them all into marks.

One page per child, in your school's name and logo: the total, the marks by type of question beside the class average, and every chapter sorted into needs work (below half), can improve and strong in, each with its marks, so a family can see why. It lists where marks were lost and why, has room for the teacher's remark - which the AI can draft and the teacher edits and saves - signature lines for the class teacher, principal and parent, and a QR code that opens the marks online once the student signs in. The chapter of each question is suggested by the AI and can be changed by the teacher. The whole class prints in one go, one A4 page per child. It gives no rank, no percentile and no pass mark.

Yes. Tick which questions on a paper are competency-based - case studies, sources, real-world scenarios - or let it read a range the paper already declares, and see the split by question count and by marks side by side, because a short competency question and a long analytical one are not worth the same. It reports your paper's own mix; it does not tell you what CBSE's target should be, which changes and is worth confirming from the board's current circular rather than from any vendor's page, ours included.

Pricing is quoted per institution after a demonstration, based on the number of students and how often you examine. The first mock test for a new institution is run free of charge.

Yes. Students write on paper as usual, photograph each sheet with their phone during a supervised upload window, and an examiner marks the pages onscreen. Across more than a hundred real question papers we studied, only a small fraction were multiple choice - the rest needed derivations, numerical working or labelled diagrams.

No. You upload the paper your teachers already wrote as a PDF or Word file. ProctorSys recovers the questions, numbering and marks, then shows them to a human for correction before the exam goes live. Papers can also be built by hand.

It assembles one from questions it has already read off your own papers. You ask for the shape you want — "Part A, ten questions worth two marks from Physics" — and it draws them from every paper already parsed for that course, spread across them rather than reusing one wholesale, then opens the result for you to correct. It does not invent questions: every one was written by your teachers and printed on a paper you uploaded. That is also the limit — it can only offer what you have already put in, so it is useful from your second term rather than your first, and nothing tracks which questions you have set before, so check the result against last term's paper.

No. Detection runs inside the student's own browser and only the name of a condition and the time it held are sent to the server. No video or audio is recorded or transmitted at any point. When a serious flag fires, a single still photograph from that moment is kept as evidence.

Yes, and this is the main reason to choose it. Because no video is ever streamed, ProctorSys needs a fraction of the bandwidth of camera-streaming proctoring systems. Only short text flags travel to the server.

Thirty days, then deleted automatically. If a student was warned or blocked during that sitting the photographs are kept for one year instead, so a disputed decision can still be reviewed. Identity photographs are deleted at the end of each session.

Yes, using the method NBA itself publishes - the share of students who crossed a threshold on each question, averaged over the questions carrying an outcome, never a plain average of marks. It works even if the exam was sat on paper: upload the question paper and your existing marks grid and get the attainment sheet back without running anything through ProctorSys. It computes per-assessment attainment; CO-to-PO rollup, exit surveys and consolidation across a whole programme are your OBE workbook's job, not ours.