Selling in dozens of countries means a customer can write in any language — and a message in Japanese or German with no agent free to read it sat for days. We put a secure, real-time translator inside the agent's screen, so anyone could answer anyone.
Expanding into international markets brings a hard operational problem with it: the language gap. This company was paying for it in swollen costs and a scramble to hire and keep agents for every language it sold in.
It ran separate, dedicated teams per language, which was deeply inefficient — one team sat idle while another, in a different language, buckled under load. And a customer who wrote in a less common language could wait hours or days for a speaker to free up, which bled into slow handling and cancelled orders.
Worst of all, agents trying to help reached for free, external browser translation tools — exposing customers' personal details to outside servers and producing garbled, embarrassing replies. The workaround was itself a risk.
Two-way, real-time, on-brand — and every byte of personal data stays inside.
We connected a secure, multilingual AI engine directly into the agents' existing omnichannel workspace — no separate tool, no copy-pasting into a browser tab.
When a customer writes in a foreign language, it appears to the agent instantly in their own; the agent writes an ordinary reply, and the system renders it into the customer's language in a fraction of a second, tuned to the brand's terminology.
A security protocol keeps all customer personal data translated and processed inside the organisation's own closed, controlled cloud environment — no leakage to external servers, unlike the free tools agents had been using.
Translation is matched to the brand's professional and product terminology, so replies read as a fluent native speaker wrote them — not as a machine guessed.
Because translation is built in, a ticket in any language can go to any available agent — ending the idle-team-here, overloaded-team-there imbalance that separate language desks created.
The AI translates; the agent understands, judges and replies in their own words. The customer relationship stays human on both ends of the language.
The language desks came down, and the wait times with them.
Every agent in the centre effectively became a “global agent,” able to handle tickets in more than 20 languages from the same screen — and first response time for uncommon languages fell from about two days to minutes.
Cutting the need for isolated, expensive per-language teams saved recruitment and operating costs by double-digit percentages, while international-customer satisfaction rose by around 35%.
The reason agents adopted it — and the reason security signed off — is that the translation lives inside the workspace with personal data kept in the organisation's own environment, not pasted into a free tool on the open web. It's the same in-workspace, human-in-control pattern as our agent-assist work: the AI removes the friction on the agent's screen, and the person still owns the conversation.
Reef TRH