You already interviewed that candidate. Three weeks ago, in the two minute voice note nobody turned into text.
Tuesday, 11:40. The admin assistant opening has been live since Thursday and the HR number has twenty three unread messages, nine of them voice notes. In the one that runs 2 min 14 s, a candidate says which neighbourhood she lives in, that her notice period is fifteen days, that the salary range works for her and that she is halfway through a payroll course. You listen, you decide she is a fit, and you move to the next one without writing a single line.
Illustrative scene, built from what repeats in hiring for frontline roles.
On Friday the hiring manager asks who the three finalists are. You do not have three names. You have nine voice notes and a memory. So the process starts over: listening to everything again, or booking interviews with people who already answered the questions.
The claim here is one you can argue with: a candidate voice note is not a message, it is the first interview. Treat it as a message and you pay twice, because you listen now and you listen again later.
Where one hour per opening goes
Back of the envelope, with the assumptions on the table so you can swap in your own:
- 25 candidates send a voice note over the life of one opening
- 1 min 50 s on average, which is 45 minutes and 50 seconds of first listening
- 10 replays to compare people and write the replies, another 18 minutes
- 1 hour and 4 minutes per opening, before a single real interview
Twelve openings in a month come to 12 hours and 48 minutes. A day and a half of work spent listening to people talk, with nothing written down at the end of it.
To price that, use a public reference instead of a number out of thin air. In Brazil, where WhatsApp carries most frontline hiring, the average starting salary in April 2026 was R$ 2,386.56, according to the Novo Caged executive summary from the Ministry of Labour and Employment. Over 220 monthly hours that is R$ 10.85 an hour, so twelve openings cost roughly R$ 139 in listening.
And that is where the arithmetic misleads. The expensive loss is not the R$ 139. It is the candidate who never got an answer, because nobody could remember what she said.
What exists today, and where each option stops
| Option | What it solves | Where it stops |
|---|---|---|
| Playing at 1.5x or 2x | Cuts about a third of the listening time | Nothing ends up written, so you cannot search it or compare two people |
| Native WhatsApp transcripts | Text without leaving the app, generated on the device | One message at a time, by long press, and few languages at launch |
| Asking candidates to type | Text at the source, in the shape you need | Many of them will not switch format, and first contact slips |
| Typing the key points yourself | A record in the right place, in your own words | The time doubles, because it is listening and then writing |
| Automatic transcription on WhatsApp Web (ZapVox) | The text is already there when you open the chat | Needs the computer on, and comes with a daily quota per plan |
Native transcripts are the option most recruiters do not know about. WhatsApp announced them in November 2024: you long press the message, tap transcribe, and the text appears, generated on your own device, according to the official blog. That solves one stray voice note. It does not solve a queue of nine.
The queue every hire leaves behind
The scale of this is not a matter of opinion. The Gupy job portal in Brazil held 6,114,199 registered candidates, published 1.4 million positions and recorded around 1.02 million hires between July 2024 and June 2025, according to the Brazilian Labour Market 2026 report.
Every one of those hires left a queue of people waiting for one sentence. The people in that queue remember the silence, and they tell others about it.
What the voice note has that the CV does not
A CV is written to pass screening. A voice note is spoken without editing, which is why it carries what actually decides the shortlist: whether the person can explain what she used to do, whether the notice period fits your deadline, whether the pay expectation matches the ad, whether she has worked with the system you run.
For frontline roles those four points screen better than any PDF layout. They arrive as audio because the candidate is on the street, on a break, with one hand busy. The format is hers, not yours.
Treating it as a message throws away the best information in the process. It is also why so many people complain about never hearing back: HR kept nothing to answer with.
Four steps that work before you install anything
- Three questions in the ad itself. In the posting and in your first reply, ask for the three facts that decide screening: city or neighbourhood, pay expectation, and when they can start. Voice notes still count. They just have an agenda now.
- Two cut off times. Listen to candidate audio at 9:00 and 16:30, in a block. Outside those windows nobody opens the inbox, because screening done ten minutes at a time never finishes.
- One line per candidate, in one place. Listened, written: name, role, the three facts, the decision in a word. A spreadsheet is enough. What matters is that the record lives outside your head.
- Two saved reply templates. One for people moving forward, one for people who are not, for now. With the line from step 3 in hand, each reply takes forty seconds, and the queue closes on Friday.
That method is entirely manual and it already removes half the problem. What is left is step 3, the one nobody keeps when the day gets tight: turning speech into text.
Where ZapVox shortens each step
With the extension open on WhatsApp Web, Automatic transcription delivers each voice note already as text, with no click, in seconds.
For step 1, Text shortcuts expand your three questions when you type the shortcut and press Tab. For step 2, Bullet-point summary turns a four minute voice note into a few lines, and Per-chat settings, under the gear icon, keeps each opening on its own configuration.
For step 3, Notes per contact keeps your line beside the chat, and Search inside voice messages finds the spoken word when the hiring manager asks which of them accepted the weekend shift. At the end, Export the chat transcriptions pulls everything out as .txt, and Glossary and names your way teaches the spelling of names so transcripts do not invent surnames.
Want to try it on your company WhatsApp today? It is 3 days free, no card.
Where this stops
Automatic transcription lives in the browser on a computer. If the machine is off or WhatsApp Web is closed, the audio sits there waiting, and on the phone you are back to the native long press.
There is a quota. The Free plan covers 8 transcriptions a day and up to 15 minutes of audio, Starter 60 a day and up to 30 minutes, Pro 150 a day and up to 150 minutes, as listed on the plans page. For an opening with 25 voice notes of nearly two minutes each, Free runs out before lunch.
Audio recorded on a busy street, with two people talking over each other, comes out worse. Place names and surnames go wrong more often than anything else, which is why the glossary is worth setting up before a high volume opening.
And the limit that matters most: a transcript is a working tool, not a document with evidentiary weight, and it does not decide who moves forward. All it guarantees is that the decision gets made by reading what the person said, instead of half remembering it.
The replies you owe nine people
Open the oldest role you still have live and count the unheard voice notes in it. That number is your reply backlog, and it only grows while the queue is listened to ten minutes at a time.
Installing it on Chrome takes less time than listening to one two minute voice note. The trial runs 3 days with everything open, no credit card. When it ends the account drops back to the Free plan on its own and nothing is charged.