Nobody at HKR scores applicant tests by hand. Minutes after a test arrives, at night and at weekends too, the grammar is marked against a fixed key and AI scores the writing against a rubric. The test has a fixed pass mark. A candidate below it gets the reason and can ask a recruiter to review the test, and a recruiter decides who is hired.

Every five minutes, day and night, the AI looks for new tests. The grammar, vocabulary and reading answers are marked against a fixed answer key. AI scores the two writing tasks against a rubric, and each test gets a language level. A candidate passes only when both parts clear the bar. About four tests in ten arrive outside weekday office hours, and they are scored just as fast.

The scores go into the hiring system as a note on the candidate. A candidate who clears the pass mark moves to the next stage. A candidate below it gets the reason and can ask a recruiter to review the test. The recruiter talks to the candidates who move on and decides who is hired.

Nobody at HKR reads every test for signs of AI-written answers. The AI holds back each test that shows those signs, with no stage move and no archive, and sends it to a recruiter. The recruiter decides what happens to that candidate.

While it scores a test, the AI also looks for signs that the answers were written by AI. A test that shows them is held: the candidate is not moved on and not archived, and the note in the hiring system says the test waits for a person.

The recruiter gets the held test with its scores, reads it and decides what happens to the candidate. Nothing about that candidate changes until then.

Nobody at HKR opens the test sheet, the hiring system and the interview calendar each morning to see that they work. At 9:00, the AI tests all three and tells the recruiter when one is broken. A person still decides how to fix it.

Every day at 9:00, the AI opens the test sheet, the hiring system and the link candidates use to book an interview, and checks that each one answers. It also checks that the night’s tests were scored and that no update to the hiring system is still waiting.

When everything works, it stays quiet. When a tool is broken, the recruiter gets one message that names the tool and what failed. A person decides how to fix it.

Nobody at HKR searches the hiring system to answer a question about one candidate. A recruiter asks in chat and gets the test, the feedback, the interviewers and the offer status in one reply. The recruiter still decides the next step.

A recruiter asks about a candidate in Slack, by name, in plain words. The AI reads the candidate’s record in the hiring system and replies in the same thread: the test result, the interview feedback, who interviewed the candidate and where the offer stands.

A lookup only reads the record. The recruiter decides the next step.