Bryan Murphy, CEO of Smartling, gave an engaging presentation during SlatorCon London on May 23, 2024 titled “Cracking the Code: Advancing Localization with Effective AI Strategies.” Murphy outlined for attendees the way Smartling approaches the entire localization process, the company’s latest technology developments, and where AI comes into play in all of it.
Murphy explained first that he approaches translation and localization from the perspective of a buyer. He focuses on the needs of clients, asking them about their problems, which, he added, have lately revolved around using AI to optimize translation efficiency. The CEO acknowledged that clients are being pressured to use AI while at the same time they express distrust in it, due to its novelty.
“It’s very new technology. And so when I run into situations where I don’t trust something, a new technology, I rely on the data,” said Murphy, adding that it is important to work with the right talent and experts to navigate those new technologies.
The CEO told attendees that over the past two years Smartling has been building out frameworks to assess the impact of AI and figure out what works and what doesn’t work. The goal behind the effort, explained Murphy, is to accelerate customer growth through translation innovation with technology at the core.
“We’ve been a cloud-native company for over a decade. We were very early on developing machine learning and neural machine translation technologies. And so when GenAI came along, it was sort of natural,” said Murphy as he described Smartling’s massive investments in R&D, particularly around GenAI.
Murphy emphasized the importance of recognizing and adapting to technological advancements within the language industry to avoid becoming obsolete. Posing the question of whether AI can help deliver the best quality, speed, and experience as pressure increases on customers to cut costs, Murphy shared some examples and results with attendees.
After giving a brief introduction to AI, including terminology, the different layers of the AI spectrum, and the various types of LLMs, their capabilities, strengths and weaknesses, Murphy addressed the question of quality. Comparing making “a gajillion widgets a day” to the need to apply quality control in high-volume translation, he shared his belief that quality requires a standardized, non-subjective way to measure it and communicate it to stakeholders.
Murphy also drew a parallel between the translation industry and the data storage/cloud services model, suggesting that as value increases, prices decrease, but volume increases significantly. To him, the focus should be on creating a solution to meet that demand, with different types of workflows that combine human and machine translation and assure high quality.
“I think that … there’s probably an infinite amount of content to be translated, being generated every day. So I don’t think we have a demand side problem to come up with a really efficient supply side solution.”
I don’t think we have a demand side problem to come up with a really efficient supply side solution.
Bryan Murphy, Smartling CEO
Smartling has rolled out a new AI toolkit, and Murphy shared with Slator that in the past year, the combination of AI-enabled translation and human translation, or “AIHT,” has resulted in a reduction of about 50% in costs and has doubled the speed of delivery to clients. The tool offers access not just to Smartling’s own operatives, but also to other LSPs, commented Murphy.
Customer-driven technology integrations are also part of the company’s strategy, but as integrations continuously evolve, they also create a challenge, explained Murphy. He highlighted automation as one of the main benefits of the Smartling translation platform in this regard, which is connected to the technology stacks that clients manage for content and other business functions (via APIs or connectors).
