A retake policy is the set of rules that govern what happens when someone fails a test or exam: how long they wait before trying again, how many attempts they get, what they must do in between, and what they may or may not do at work meanwhile. For roles that oversee AI agents, the last point matters most.
Why does a retake policy matter?
Without one, failures are handled case by case. One manager lets a reviewer resit the next morning; another makes them wait a month. Approval rights end up depending on who your manager is. A written policy makes the outcome the same for everyone.
It also protects the meaning of the pass. If people can resit immediately and as often as they like, the test becomes a guessing exercise. If the next attempt draws different questions and requires some learning first, a pass shows the person has closed the gap.
What does a retake policy cover?
A workable policy for agent oversight answers six questions:
| Question | Typical answer |
|---|---|
| How soon can someone resit? | After a set gap, long enough to study the weak areas |
| What must happen first? | Review the topics failed, with a mentor or set material |
| Are the questions different? | Yes, drawn from the same plan but a different set |
| How many attempts? | A fixed number before a fuller retraining step |
| What can the person approve meanwhile? | Nothing they would need the certificate for |
| Who decides exceptions? | A named owner, recorded |
The fifth row is where agent work differs from ordinary training. If a current certificate is a condition for approving held actions at a given depth, a failed test means the person stays off that queue until they pass. That should be stated plainly so team leads do not improvise.
An example
A lender requires reviewers of its affordability agent to pass an internal test of 30 questions, with a pass mark of 24 correct. A reviewer scores 21, missing four of five questions on declared income. Under the retake policy, she waits two weeks, works through the income verification material with a mentor, and resits a different set. In the meantime, she reviews only drafts at depth 2 that do not need the certificate. Her second attempt scores 27, with all income questions correct.
How is a retake policy different from an AI policy?
An AI policy sets the organisation's overall rules for using AI: what is allowed, who owns it, how risks are managed. A retake policy is a narrow procedural rule inside the training regime. It often sits as a clause within an AI literacy policy, alongside who must be trained, how competence is tested and when certificates must be renewed.
The AI agent policy template and the guide to writing an AI agent governance policy show where training rules fit in the wider set.
Which terms are related?
- Pass mark: the threshold the retake policy responds to.
- Question rotation: what makes a resit a fair new test.
- Recertification: renewal of a certificate, with its own rules on lapse.
For the wider training approach, read training teams for AI agents.
Where the Academy fits
In Pryme Intelligence Academy, lessons, labs and tests are free. Professional certificates carry an issue fee, paid to be issued a certificate and never to pass, so a resit is a matter of learning, not cost. A business can add its own rule on what a person may approve until they hold a current certificate.
Questions people ask
Should a retake policy limit the number of attempts?
Yes, with a fuller retraining step after the limit rather than a permanent bar. Repeated failure usually means the person needs more practice, not more attempts.
Does a retake policy apply to renewals?
It should. If someone fails a renewal quiz, the policy should say whether their current certificate stands until expiry or is suspended at once.