A proctored exam is one taken under supervision, in person or remotely, so the issuer can confirm who sat it and that the rules were followed. The proctor checks identity, watches for help that is not allowed, and records anything unusual. Proctoring does not judge the answers; it protects the claim that the named person earned the result alone.
Why does proctoring matter for AI agent credentials?
Some credentials carry real authority. A person certified to build and certify agents shapes what those agents may do for customers: their brief, their sources, their bounds. If the exam behind that credential could be sat by someone else, or completed with a colleague's help, every agent the holder builds rests on a weaker claim than it appears.
Proctoring also matters to the people who rely on the credential later. A client, supervisor or employer checking a certificate wants to know the result belongs to the person named on it.
How does a proctored exam work in practice?
Remote proctoring is common and usually involves:
- identity checks against a photo document before the exam starts;
- a check of the room and desk through the candidate's camera;
- continuous video and screen monitoring during the exam;
- rules on what may be open, such as product documentation or nothing at all;
- a record of any flagged events, reviewed before the result is confirmed.
In-person proctoring does the same in a test centre. Either way, the exam itself still has to test the right things. Proctoring a poor exam produces a well-guarded poor result; what to test is covered in what an AI agent exam should test.
What is an example?
A candidate for a builder credential sits a remote exam on designing and certifying agents. The proctor checks their identity, confirms the desk is clear, and monitors throughout. Mid-exam, the candidate's screen shows a second window opening. The proctor flags it; on review, it was a system notification, closed at once. The result stands, and the flag and its resolution are on file. If a client later questions the credential, the record shows how it was earned.
How is a proctored exam different from an open-book exam?
They answer different questions. Proctoring is about supervision: who sat it and whether the rules were followed. Open-book is about resources: whether candidates may consult materials during the exam. An exam can be both proctored and open-book, with the proctor confirming that only the permitted materials were used. The trade-offs of allowing materials are set out in what an open-book exam is.
For organisations writing their own exams, the structure comes first: what an exam blueprint is, how to design an internal AI agent exam and the internal exam blueprint template.
Which terms are related?
- Proctor or invigilator: the person who supervises the exam.
- Assessor: the person who judges evidence, a different role.
- Capstone: practical work that often accompanies a proctored exam.
For the wider topic, read AI literacy and certification for people who work with agents.
Where the Academy fits
Certified Pryme Builder in Pryme Intelligence Academy requires a proctored exam and a capstone agent certified in Attest to depth 2 or higher. The exam confirms knowledge under supervision; the capstone shows the candidate can produce an agent that clears certification on real cases.
Questions people ask
Is remote proctoring as reliable as a test centre?
It can be, if identity checks, room checks and monitoring are done properly and flags are reviewed. The weak point is usually the review of flags, not the technology.
Do all AI agent exams need proctoring?
No. Low-stakes knowledge checks, such as module tests during learning, do not. Exams behind credentials that grant authority to build or approve should be proctored or backed by practical evidence that is hard to fake.