Generative AI, a.k.a. GenAI, has made the leap from digital experimentation “sandboxes” to daily use and C-suite plans for the year ahead. But beyond the point of pride that comes with displaying “AI” prominently on a website, the question remains: How can AI — and more specifically, large language models (LLMs) and GenAI — translate to ROI?
That was the question posed during SlatorCon Remote March 2025 to panelists Simone Bohnenberger-Rich, Chief Product Officer at Phrase; Simon Koranter, Head of Global Production and Engineering at Compass Languages; and Matteo Nonne, Localization Program Manager at sportswear brand On.
Beyond the hype, Slator Head of Research and panel moderator Anna Wyndham asked, how is GenAI transforming localization’s role in business — particularly the perception of localization as a growth driver versus cost center?
“I think LLMs are possibly a once-in-a-lifetime opportunity for localization teams to move much closer to that revenue-generating story, and we do see that,” Bohnenberger-Rich said. Some localization customers are already managing to do that, she noted, while others are moving in that direction.
Context-rich LLMs, which produce nuanced outcomes and results, are well-suited to help localization teams shift into the business of content adaptation and transcreation, which really drive and maximize engagement with target audiences. “That’s where the money is,” Bohnenberger-Rich added.
Koranter pointed out that success depends in large part on clear communication in managing clients’ expectations, which generally fall into one of two categories.
“Some clients were overexcited about the opportunity, possibilities, and potentials of GenAI — ‘I’m going to press a button and all my worries will go away,’” he said. On the other hand, other clients looked at developments “with a bit of fear in their eyes, thinking this is sort of like a ‘Skynet’-type scenario.”
The localization team’s role, Koranter explained, is to bring both types of clients an understanding of what the technology — and localization professionals — can do, thus managing their expectations.
“We have managed to show them it’s vital to put experts in the loop, otherwise the business objectives might not happen,” Koranter said. “When we build custom solutions, there’s a mutual understanding of the AI being a really valuable tool rather than a wholesome and complete solution.”
Nonne agreed that Koranter’s description — of some seeing AI as a cost-saver, with others recognizing greater potential for strategic innovation and a more tailored approach — could be seen as a “microcosm of what’s happening around the world, and perception in general.”
“I think […] localization leaders are in the best position to educate others on the topic because we’ve been exposed to AI for quite some time now,” Nonne added.
Putting GenAI to Work
Localization teams and leaders alike can talk about GenAI as a revenue driver, but clients, understandably, want to see proof.
For Bohnenberger-Rich, one of the most promising use cases is “hyper-personalization”: dynamically generating and adjusting multiple versions of content based on demographic data, including locale and gender.
“We know that 70% of consumers expect personalized content. We know that companies that excel at personalization have 40% more revenue than their peers. And we know that companies who excel at personalized content also have twice the conversion rate than others,” Bohnenberger-Rich said.
“I think LLMs are possibly a once-in-a-lifetime opportunity for localization teams to move much closer to that revenue-generating story,” — Simone Bohnenberger-Rich, Chief Product Officer, Phrase
When it comes to automated customization, not all content is created equal. Nonne shared that at On, “dry” content — very factual elements such as technology sheets and legal contracts — is up first for localization.
On has recently expanded to produce content for certain marketing channels, having trained LLMs with past localized content. This allows the company to deliver content a bit closer to the tone of voice and expected outputs.
On decides when to prioritize a human touch based on the impact specific content has on customers as a main target audience.
“It’s about defining where humans [add] the most value and where they can craft messaging that really drives the point home with international audiences, and where there’s high visibility for them, versus what doesn’t need the same level of attention,” Nonne explained, with AI being leveraged more frequently for the latter category.
ROI: Beyond a Reduction in Cost
According to Koranter, clients’ motivation for introducing AI into workflows is typically not restricted to cost-saving
“It’s also not a sprint,” Koranter said. Continuous localization projects and workflows take time — time to make sure the solutions in place make sense and deliver the needed results; time to meet clients’ various expectations and business objectives; and time to bring to life clients’ bespoke solutions.
“We have managed to show them it’s vital to put experts in the loop, otherwise the business objectives might not happen,” — Simon Koranter, Head of Global Production and Engineering, Compass Languages
Starting with a clear understanding of client expectations and goals, companies can educate on what’s available, what’s on the horizon, and what can be tested out. Those factors then contribute to the end-product and extensive workflows, in order to keep content fresh and relevant, to ensure customer engagement and business results.
“The main ROI, when it comes to the bottom line for the clients, might not be a reduction of cost […] but mostly faster turnaround times, better impact, better results,” Koranter said.
“It really depends on the business context of our stakeholders,” Bohnenberger-Rich agreed.
For a cryptocurrency trading platform, the priority might be time to market and achieving “localization on day one” across all assets instantaneously.
Large retailers under tremendous competitive pressure often focus on cost, but that term can cover many factors, from stretching a localization budget to driving up engagement via apps and digital assets to reducing human intervention, as measured by QPS and intelligent routing.
“Some of our customers see a 50-70% reduction in human review effort, and that clearly hits the bottom line,” Bohnenberger-Rich said. “It’s a cost.”
The More Things Change…
Priorities and goals can, of course, change as technology evolves — and with GenAI developing so rapidly, planning for the year ahead very well might incorporate those advancements.
“I think […] localization leaders are in the best position to educate others on the topic because we’ve been exposed to AI for quite some time now,” — Matteo Nonne, Localization Program Manager, On
Nonne said he is particularly excited about AI agents, which could add a layer of quality control to localization that can further increase efficiency without necessarily compromising quality. Tied to that, he noted, he foresees the industry moving away from its traditional English-first approach, instead looking at content creation in multiple languages at the same time.
Bohnenberger-Rich said she believes the most important shift in the next months and quarters will be a renewed focus on data assets and models.
“I think localization teams sit on a treasure trove of data which is key to that personalization story,” Bohnenberger-Rich explained. “LLMs can only be customized if you target them and infuse your own data, and that data has to be in top-notch shape and form.”
That means not throwing the baby out with the bathwater, Koranter pointed out. Translation memories (TMs), he said, are still a vital linguistic asset for production that “still need to be very much in the game.”
“One day it might be a different data asset, different structure input, but it’s highly relevant,” Bohnenberger-Rich said. “These fundamentals in AI have not changed with LLMs. Data is still king.”