Using Orova Recruit
step by step.

Let the AI write the JD. Type a few bullet points into the Job description (JD) box, then press Write JD with AI. The AI uses exactly what you typed as the brief and writes the finished description straight into the position name and the JD box, where you edit it like any text.

Or bring your own. Press Upload JD to add the file, or paste the text in. It takes PDF, Word (DOCX/DOC), TXT, RTF, ODT and JPG/PNG photos or scans, up to 10MB. A scan is read by the AI directly rather than by a text extractor.

Press Analyze JD to get the criteria. The AI splits the JD into 8 to 16 criteria across five groups — Required, Important, Preferred, Basic and Flexible — each with a weight from 1 to 10, plus a suggested pass threshold. Each criterion is then broken into 2 to 5 small checks, and those checks are what a CV is actually scored on. Add, rewrite, reweight or delete any line.

Set the pass bar. Two numbers decide the labels: the Pass threshold and the Review band. At or above the threshold is Pass, inside the band just below it is Review, anything lower is Fail. Required criteria are special: scoring under 50 on any of them marks the CV Fail however strong the total looks.

Decide who may see this opening. The Recruiters button opens the opening's own access panel: the default for workspace members is Can edit, View only or No access, and you can grant individual people their own level on top. The creator and the workspace owner always keep full access. Set it to No access and the opening disappears from every list except theirs and those you named.

Deleting is reversible. Deleting an opening only moves it to the trash; restore it at /recruit/trash — the trash is not on the menu, so type the address. Deleting an individual CV, on the other hand, deletes the file for good.

Upload in bulk. Press Upload CVs to pick one file or many, or Upload folder to send a whole folder. Up to 500 CVs per opening, 10MB each, in the same formats the JD accepts. The browser sends 10 files per batch and the progress card counts the real bytes.

Score one, some, or everything. Press Analyze all, or tick several rows and open the Actions menu. Scoring runs in the background, one file after another, and the screen names the file being read and how many are done out of the total.

Every score is a list of checks. Each criterion is scored check by check: met or not met, each one carrying a quote lifted straight from the CV. The criterion score is checks met over the total. The overall score is a weighted average computed by the server — not whatever number the AI adds up.

What the AI pulls out. The candidate's full name, email, phone, years of experience and latest role, plus an overall assessment. The result panel also previews the file itself — or the text inside it when the browser can't display that format — and offers a download.

Two verdicts side by side. Next to the AI screening column sits Your call, where you set Pass, Review or Fail yourself. The two opinions stay side by side, so a disagreement is visible rather than overwritten.

Nothing is lost. A file with no extractable text and no readable image is marked Unreadable — the AI is never called and no quota is spent. If your quota runs out mid-run, the run stops and the remaining CVs stay Not analyzed.

One shared set per opening. Pick Shared set (from JD) in the candidate dropdown and press Generate. That set is the default every candidate inherits, so a whole panel of interviewers works from the same list.

Or a set written for one candidate. Pick a scored CV instead and the questions aim at that person specifically: the claims worth probing and the gaps against the JD.

Every question is also a scoring line. Each card holds the question, the interviewer's scoring criteria, the Score from 1 to 10, and the 2 to 5 checks the answer is marked against. Those checks are what turns interview answers into a comparable number later.

Edit freely. Rewrite any card, drag the handle or use the arrows to reorder, add one from the dashed row, delete what you don't need. Cards save themselves when you leave them — there is no Save button to forget. The AI button on a card polishes your wording without changing your meaning.

Apply to everyone. Apply to all pushes the shared set to every candidate on the opening. Personal sets are replaced and their old reviews cleared, because the scorecard they were scored against no longer exists.

Type a line, press Enter. Notes are individual lines, like a chat. Each one saves itself and carries the time, the date and the email of whoever wrote it — useful when two interviewers share a room.

Edits stay visible. A changed line keeps a small edited marker; open it to see the current text alongside every earlier version with its timestamp.

Let the microphone write for you. Press the mic and the browser listens and types the answers as they are spoken. Pick your Listening language from 56 options — the choice is remembered — and note that live listening happens inside the browser, so it costs no quota. On a browser without live speech recognition, Orova records the session instead and the AI transcribes it when you stop.

The session is recorded too. Recording runs alongside the live dictation, so the audio is there to replay from the Review tab or download later. Near the two-hour mark the file is cut and saved on its own, and recording carries on.

Get the full transcript. Once the recording is saved you are offered Get the full transcript: live dictation can drop words at pauses, so the AI listens to the whole file and appends the complete version to your notes. This one uses quota.

Multiple rounds per candidate. The Session dropdown shows the round you are recording plus every round already saved. Save session freezes the current round — notes, question set and recording — then clears the panel for the next one. Saved rounds reopen read-only.

Score against the checks. Press Review all, or review one candidate at a time. The AI reads your notes and the recordings against the question set and marks every check met or not met with a short quote that says which source it came from. A check neither source covers is marked not met and labelled as not covered — it does not guess in anyone's favour.

The maths is the server's job. The AI hands back no number at all. A question scores checks met over total, multiplied by that question's Score; the interview score is that sum divided by the sum of the Scores, expressed out of 100.

Work through the table. Search for a candidate; expand a row to replay the recordings and read the notes with times and authors; attach a recording still sitting on your device; or tick several rows to review them in one go, set the decision, or download their reports together.

Compare the shortlist. With at least two screened candidates, AI recommendation ranks them using both the CV score and the interview score, gives its reasoning per person and flags its AI pick. It is advisory — the note beneath it says so.

Record the final call. The Final call column is Hire or No hire, pressed by someone with edit rights on the opening; pressing the same button again clears it.

Take the report away. Download PDF builds a hiring result report — the CV screening part and the interview part, with when each was scored and when the file was exported — for the whole opening or for one candidate.

Declare your rooms first. The Rooms button: each room has a name, an address or directions, a default number of minutes per slot, and its booking hours. Set one set of hours for every day, or switch to Per weekday and give each day of the week its own — including days marked Closed.

Book a slot. The Book interview button: choose the opening, the candidate, the room, the date, the start time and the duration, add a note, and pick a Card colour to tell types of interview apart at a glance.

Impossible slots are refused. Outside the room's hours, on a closed day, or clashing with another booking in the same room: the screen blocks it and the server checks again. The message names the room and its hours, or the time already taken.

Move around the calendar. Three views — Day, Week and Month — with Today and the two arrows. A full day cell shows how many more interviews it holds.

The invitation goes out on its own. Interview just scheduled is one of the conditions the automatic email rules listen for, so the invitation can leave the moment the slot is saved. You set that up under Contact.

Connect your mailbox. On the Email account tab: the Connect Gmail card fills in Gmail's server and opens the page for a 16-character app password, or you type your own SMTP host, port, security and sender display name. The password is encrypted before it is stored, and Save & test runs a live login so you find out immediately if it is wrong. Only the workspace owner and admins can set this up.

Write your templates. Keep as many templates as you like; the list shows the name, the Template ID, the subject, who created it and when it last changed. In the editor you write the template HTML, use Insert variable, use Insert image (JPG, PNG, WEBP or GIF, up to 3MB), or ask the AI to generate the template and keep chatting so it edits the existing one rather than starting over.

Let the variables fill themselves. The variable menu is grouped: basic info covers name, email, phone number, company name and job title; CV screening covers the AI screening result, your call and the CV score; interview covers the AI verdict, the interview score and the final call; scheduling covers the date, start time, duration, room, room address and note; plus the per-candidate message body.

Check it on three screen sizes. The preview shows the email as Desktop, Tablet and Phone before anything goes out.

Send automatically on status. On the Send email tab, each rule reads this condition sends this template. There are five conditions: AI screening result, Your call on the CV, AI interview verdict, Final call, and Interview just scheduled. Each rule sends only once per candidate — a log prevents duplicates — and any rule can be switched off. A small panel counts your templates, active rules and emails already sent automatically. This is how rejected candidates still get a courteous reply without anyone typing it.

Free. Creating openings, uploading CVs and JD files, editing criteria and questions by hand, booking interviews, downloading the PDF report, and the live microphone dictation inside your browser.

Charged. Analyzing a JD and splitting the criteria into checks, writing a JD with AI, scoring a CV, generating or polishing a question set, reviewing an interview, the AI candidate comparison, an AI email template, and AI transcription of a recording.

Nine interface languages. Vietnamese, English, Spanish, Chinese, Japanese, French, Thai, Lao and Khmer. Everyone picks their own.

One wallet. Orova Recruit draws from the same quota wallet as Orova SEO, Orova Ads and Orova Insight, and a new account starts with 1,000 quota — enough to run a real opening end to end and judge it for yourself.