Translation used to mean long turnarounds, high per-word costs, and messy quality control. AI changed the constraints.
Bryan Murphy, CEO of Smartling, explains what AI translation looks like once it moves into production and why speed alone doesn’t solve the enterprise problem. In the episode, Murphy describes the old baseline in the translation market: roughly a 14-day turnaround and about 20 cents a word, which pushed teams to translate as little as possible. He also points to the risk created by “instant” translation: if a company publishes a long asset across dozens of languages, what did it actually say in each market, and can anyone verify it—especially in regulated categories?
Murphy describes how Smartling is built: custom translation engines trained per customer on existing linguistic assets, workflow automation that integrates with enterprise systems, and human-in-the-loop review when content warrants extra oversight. He says Smartling translates eight to nine billion words a year across thousands of customers, which include OpenAI, Apple, Disney, Shopify, Pepsi, and Verizon, to name a few.
Listen to the full episode to hear Murphy’s operator view of translation, governance, and what changes when AI becomes trainable at enterprise scale.