For hiring teams
AI hiring, from the buyer’s side
This page is for hiring teams looking at AI interview services, or already using one. More than twenty are available in Japan, and their feature lists look much the same. Here you will find what each service optimises for, where the real differences lie, what to check when choosing, and how to tell afterwards whether the evaluations are right. If you only read one article, start with AI interview services compared, which sets out how the services differ.
AI interview services compared →
Features and prices look alike, so they do not settle the choice. We sort the leading services by what each one optimises for, so you can see which type fits you and what to ask when you are choosing.
Can AI evaluate candidates correctly? HR writes the criteria →
AI interview scores follow the criteria that HR writes. We cover variation between interviewers, how to compare human and AI scores, two ways to prepare the criteria, and the cycle of reviewing them in use.
Question design and AI interviewer behaviour: keep improving them from the records →
Candidate complaints are more about how the conversation flows than about what is asked. We show how to use interview records to improve the questions and the AI interviewer’s settings, and three ways to review them in Proomi.
More criteria do not mean more evidence →
If scores bunch up, extra criteria add little evidence. Using published performance-review data for Japan’s national civil servants, we show why this happens and what to check when choosing an AI interview service.
Was your interview evaluation right? You find out after the hire →
Structured interviews predict performance well on average, but far less in some workplaces. We set out why evaluations, AI interviews included, should be checked against post-hire outcomes, and what you need to do it.
Rewrite your criteria, then test them on past interviews →
An AI interview keeps the full transcript, so past interviews can be re-scored after you rewrite your criteria. We cover what this can and cannot test, and the order in which to check.
Same answer, different score: when interviewers read the criteria differently →
Even with the same criteria, scores depend on who gives them if people read the criteria differently. We look at what research shows, and how gaps between human and AI scores reveal where readings differ.
Use your performance review criteria in interviews, word for word →
Performance review criteria work as interview criteria without rewording. A persona simulation shows which criteria lack evidence and what changes when you add questions.
Choose an AI interview tool that can explain its scores →
What the AI reads, what standard it scores against, and what happens when there is no evidence: the questions to ask when choosing a service, with Proomi’s answers.