HR: are you still playing employee voice notes one by one to find out what was asked?
Tuesday, 8.47am. The HR analyst opens the company WhatsApp Web and finds nine voice notes from the night before. The first runs 4 min 12 s: a machine operator explains why he missed his shift, names the clinic, gives the date of the appointment, complains about the bus, and in the last eight seconds asks to swap Saturday. The second runs 1 min 08 s and wants to know whether the meal allowance lands before the 5th. The third runs 2 min 30 s, is mostly frustration about the rota, and the actual request arrives at the end.
None of it is urgent. All of it needs an answer. And to answer, she has to listen. On headphones, because the room is shared. Taking notes, because in three weeks someone will ask what was agreed with that operator.
That half hour shows up in no report.
The channel changed, and HR was not consulted
A note on where these numbers come from: they are Brazilian, because Brazil is where WhatsApp became the default channel for work first and where the statistics are best. If your team already treats WhatsApp the way a Brazilian factory office does, the arithmetic travels.
In ICT Enterprises 2025, run by Cetic.br/NIC.br, among companies that keep a social profile, the share using WhatsApp or Telegram rose from 74% in 2024 to 79% in 2025, and 79% of companies selling online used those apps as a channel. The survey is about sales and customer contact. In practice, though, the company number does not sort what arrives. The same channel that takes a customer order takes the sick note from the warehouse, the question about the payslip and the holiday request from someone on a day off.
People on a loading bay, on a roof, in a van or with both hands busy will not type. They record. And they record the way people speak: context first, a detour, an explanation, and the request at the end.
The arithmetic nobody times
Run it with your own numbers, but here is an illustrative version for scale. Nine voice notes a day averaging 1 min 50 s is 16 and a half minutes of listening. Listening is not the whole cost: there is the tap to open, the headphones, the rewind ten seconds because the date went past, the note about what was asked. Anyone who does this daily knows the real figure lands near double. Thirty minutes a day is two and a half hours a week. Swap nine for twenty-five, which is the inbox of a mid-sized plant, and you have lost a morning every week.
HR already knows where that time could go. A survey by LG lugar de gente with FIA Business School, run in August 2026 with 1,130 HR professionals across 26 states, measured 5.4 hours a week handed back by using artificial intelligence, which is 12% of the working week. Ninety-four per cent say they already use AI in some form. The detail that matters here is in the list of uses: 38% have put a chatbot or an agent in front of their own employees.
That is the knot. HR automated the written half of internal service, and the spoken half still lands in one person's inbox. A voice note does not enter an automated flow because it is not text. Before any automation can touch a recording, somebody has to turn voice into words.
What exists today, and where each option stops
Five options HR tries, in the order they usually appear.
| Option | What it fixes | Where it stops |
|---|---|---|
| Playing at 1.5x or 2x | Cuts a third off the listening time, with nothing to install | You still listen to all of it. And names, dates and numbers are exactly what gets lost at speed |
| WhatsApp's own transcripts | Text from the recording at no cost, generated on the device | They were announced in November 2024 for the phone, with a long press on each message. That is not WhatsApp Web on the computer, which is where HR actually works |
| Asking people to write instead | Genuinely solves it, when it holds | Someone in gloves, at the wheel or on the line records again the following Monday. And the request does not wait for the next team meeting |
| Typing it out yourself | A faithful record, in your own words | A 4-minute recording costs 12 to 16 minutes of typing, illustratively. Multiply by nine and the morning is gone |
| A chatbot on the HR line | Handles what is repetitive and written: payslips, holiday balance, benefits | It works on text. The recording does not enter the flow until it becomes words, and then it is back in somebody's inbox |
None of these is wrong. What they share is that they all treat voice as the exception. For the HR inbox, it is not the exception.
How to do it in WhatsApp Web, step by step
ZapVox is a Chrome extension that works on top of WhatsApp Web. You install it from the Chrome Web Store, open WhatsApp Web and transcribe the first recording from the microphone icon. From then on the order of the routine changes: the text arrives before the decision to listen.
- Automatic transcription. New voice notes appear with the text below the message, with no click. The analyst at 8.47am opens the chat and reads.
- Bullet-point summary. For the 4 min 12 s recording, the summary gives you the lines that matter. The shift swap leaves the last spoken paragraph and becomes an item.
- Sentiment reading. Flags the tone of a message, such as worried, satisfied or in a hurry. Useful for choosing which of the nine chats you answer first.
- Glossary and names your way. You teach it the words your company uses: the department, the nickname for the night shift, the benefit acronym, the supervisors' names. The transcript stops guessing.
- Notes per contact. A private note pinned to the conversation, visible only in your browser. That is where what you agreed goes, so next week you still have it.
- Search inside voice messages. Search for sick note or shift swap and see the chats grouped by person, instead of scrolling back through March.
- Text shortcuts. The reply you type fifteen times a week becomes a slash command plus Tab, with the name filled in.
- Schedule messages and files. The answer you write at 10pm goes out at 8.30 the next morning, with the recipient verified before it sends.
- Export the chat transcriptions. Produces a .txt of the conversation, for when a case has to go to whoever decides.
- Long audios, whole. The 40-minute recording of a shift handover is split internally and comes back whole, with nothing cut off the end.
Where this stops
The part that rarely makes it onto a product page.
It is a browser extension. It needs the computer on, Chrome open and the WhatsApp Web tab live. If HR answers from a phone while walking the factory floor, this does not fix the factory floor. It fixes the desk.
There is a daily quota. On Free you get 8 transcriptions a day and recordings up to 15 minutes, search covers the last 7 days, and scheduling is capped at 3 a day. Translation, bullet-point summary, sentiment reading, sending the transcription back and .txt export sit in Starter, Pro and Trial. An inbox with twenty-five voice notes a day does not fit inside Free, and it is fairer to say so now than to let you find out on Wednesday.
It does not talk to your systems. There is no automatic link to payroll, time tracking or an HRIS. The text comes out as .txt and somebody still pastes it where it belongs. If your routine requires the request inside a system, that path still runs through you. What changes is that you get there already knowing what was asked, without having spent four minutes finding out.
Bad audio stays bad audio. An echoing warehouse, wind on the microphone, two people talking over each other, a phone in a pocket. The glossary helps, but nothing invents what was never audibly said.
It does not transcribe calls. Voice messages and audio files, yes. WhatsApp voice or video calls, no.
And it is not evidence. The transcript is a working tool: it lets you answer, pass things on and find them again. What counts formally is your own internal procedure, with the form and the acknowledgement you already use. Treating the text as an official document creates a new problem instead of solving the old one.
If you want to test it on a real week
Do not test with a recording you picked. Test with a Monday, which is when the HR inbox is fullest. Install it, leave the tab open, and at the end of the day compare two things: how many voice notes you read without needing to listen, and how many times you went back to check a date.
Every sign-up starts with 7 days of trial, with no credit card. The trial activates by itself on first login, and when it ends the account returns to the Free plan. Nothing is charged and there is nothing to cancel. The numbers and the plans are on the plans page, and the full list of what each plan includes is under features.
In two lines
Employee voice is not going to shrink, and the chatbot that 38% of HR teams have already installed works on text. Turning voice into words is the step that is missing before anything else, and it takes seconds when you are not the one doing it.