Support teams that stopped hiring for every new language
Adding a language to your support coverage doesn't mean hiring for that language. Here's what happens to ticket deflection rate, CSAT, and agent handle time when teams route by intent instead of by language.
Trakpak handles shipments across Brazil, Mexico, Colombia, and Argentina. At peak season, 60% of their support tickets came in Spanish or Portuguese — but their team was English-primary. Night shift coverage for LATAM time zones required three dedicated agents who spent most of their time on order status queries.
"We cut LATAM night shift staffing by two-thirds in six weeks. The Spanish routing quality wasn't just passable — our agents stopped checking the AI's work after the second week."
Nuvara serves customers in the UAE, Saudi Arabia, Egypt, and Morocco. Arabic-language tickets were their biggest pain point: every tool they evaluated either failed RTL rendering, misclassified intent, or couldn't handle dialect variation between Gulf Arabic and Egyptian Arabic for the same financial query type.
"Arabic was the test. Every platform we evaluated failed it. Queryvine was the first that handled it correctly — including the difference between Gulf and Egyptian dialects."
Helios operates marketplaces in Thailand, Indonesia, Vietnam, and the Philippines. Thai word-boundary tokenization failures in most NLP tools caused intent misclassification on routine queries. Their previous tool read "I want to return" in Thai as a completely different intent class 30% of the time.
"We were running five micro-teams sorted by language. Now we have one ops team with routing. The context card tells agents what they need — the customer's language doesn't matter anymore."
Orbient handles booking support across 12 European markets. Peak summer volume — when French, German, Italian, Dutch, and Polish customers filed support requests simultaneously around high-demand routes — overwhelmed their team's language coverage. Cancellation and rebooking queries were the highest volume and highest-stress category.
"Our agents used to spend 4 minutes reading a machine-translated ticket. Now the context card arrives with the escalation. Escalated ticket handle time dropped 35%."