When clients write in a language the agent doesn't know, WhatsApp's built-in translation won't save you: it only works in the mobile app, and inside the CRM the agent sees the message as-is, in a foreign language. That's why translating conversations in Kommo has to be solved separately. Below is why that happens, what the options are, and what each one "costs" in speed and errors.
- WhatsApp's built-in translation works on the phone, but not in the CRM — inside Kommo the agent sees the original language.
- Manual workarounds (Google Translate, browser extensions) slow the reply down and breed errors; a bilingual agent is expensive and doesn't scale.
- The systematic fix is translation right in the lead card: the agent writes in their language, the client reads in theirs, and the other way around.
- This is conversation translation, not a bot: a human writes the replies.
- It connects by QR in a couple of minutes, so you can install it and test it yourself.
Why WhatsApp's built-in translation doesn't work in the CRM
In the mobile WhatsApp app, messages can be translated. But the moment the conversation runs through a CRM like Kommo, that translation stops working: the agent replies in the lead card, not in the app, and sees the message there in its original language.
What comes next is what eats your conversion. The agent copies the text into a translator, translates it, writes a reply in their own language, translates that too, and pastes it back. Every single message takes a few extra steps. The client waits. And when multilingual chats are a steady flow (border markets, diaspora customers, clients abroad), this "manual" back-and-forth turns into a constant drag and a source of mistakes: wrong language, lost context, reply sent to the wrong thread.
The ways to translate conversations, and their price
| Method | Speed | Errors | Scale | Price |
|---|---|---|---|---|
| By hand (Google Translate / DeepL) | low — copy-paste on every message | high — context lost, languages mixed up | poor | "free," but the agent pays in time |
| Browser extension | medium | medium | medium | cheap, but unreliable inside the CRM interface |
| Bilingual agent | high | low | poor — you need a person per language | expensive, and you can't hire one for every language |
| Translation right in the CRM | high — right in the card | low — translates both incoming and outgoing | good | a widget; translation can be turned off |
Manual methods look free, but the agent pays — in time and errors. A bilingual agent covers one language and runs straight into hiring limits. Translation right in the CRM removes the copy-paste step and works in any language.
Translation right inside the CRM
The idea is simple: the client's incoming message is translated for the agent, and the agent's reply is translated for the client. All inside the lead card, with the conversation history kept in the CRM.
In Kommo, the AnyLinga widget covers this: it connects WhatsApp to the CRM and translates the dialogues in real time on top. It's worth being clear about what it is not — AnyLinga doesn't write replies for the agent and doesn't qualify leads. It's not a chatbot: a human writes the replies, the widget only translates. You can also turn translation off, and then it's simply a WhatsApp connector to the CRM.
It connects fast — by QR code, in a couple of minutes — so the easiest thing is to install it and test it on your own flow. One honest caveat: right now the link is by QR, meaning a personal number, so it's an "inbound" scenario (clients message first; more on this in the WhatsApp connection hub). For mass outbound broadcasts you need the official Cloud API.
Who actually needs this
Translating conversations in the CRM pays off when multilingual chats are a flow, not a rare exception:
- the business sells into border or overseas markets;
- the clients are a diaspora who write in different languages;
- an integrator is setting up a CRM for a client with multilingual support, where hiring an agent per language isn't worth it.
If there's only one language and it matches the team's, you don't need translation; a plain WhatsApp connection is enough.
What's next
Translation removes the language barrier. After that, to keep multilingual requests from going cold: first-reply speed (SLA) and distribution of inbound to a free agent stay exactly the same as for a single-language flow.