Budget pressures meant our software client needed a fresh approach to localization. But could they trust machine translation to deliver the accurate, consistent results their users were counting on? The solution was Assured AI—Rubric’s perfectly tailored balance of AI speed and human expertise.
NICE is a global software provider that helps businesses of all sizes to automate customer service, for experiences that drive customer loyalty and satisfaction. The company is at the forefront of introducing AI into contact centres, providing easy interactions and seamless journeys using its cloud CX (customer experience) platform.
Technical Publications Manager Craig Laurence is responsible for localizing NICE’s online help content, as well as elements of its UI (user interface), into 10 languages. He explains:
“When contact centre agents or supervisors are using NICE software, they need to be able to understand it in their own language. And if they have a question, it’s critical that the online help content is aligned with what they see in the UI.”
Delivering this high-quality content for all NICE’s markets is a significant piece of work: an online help release can comprise 1 million words, around 10% of which are new or changed.
The challenge: Saving costs without sacrificing quality
NICE had been partnering with Rubric for a couple of years when, in 2024, their localization budget was adjusted to meet competing costs. This meant Craig could no longer afford to rely purely on human translation.
But how could he maintain that consistent quality for a fraction of the cost?
Craig recognized that machine translation (MT) had to form part of the solution:
“We provide AI tools to our customers, so we need to walk the walk, not just talk the talk.”
However, he was concerned that real-time AI translation tools may not deliver the quality and accuracy his users rely on. For results he could trust, Craig needed an AI solution that was informed by human intelligence.
The solution: AI shaped by humans
Rubric’s Assured AI solution balances machine translation with human input, in a best-of-both approach that matches your needs and budget.
For NICE, the focus was ensuring the online help content aligned with the UI. Project Manager Sam Jermy explains how Rubric tailored the solution to achieve this:
“We identified that the UI and release notes were priority content that should always be translated by a human. We could then leverage those quality translations, and our existing translation memories, to machine-translate the online help quickly and accurately.”
We tested several different MT engines with NICE’s content to find the most effective one for each language. We compared outputs using our automated quality-checking tool, as well as consulting with linguists.
The selected MT engines were trained using previous translations and specially curated glossaries, enabling them to “learn” NICE’s content patterns and select the correct terms. Outputs were reviewed to check that terms were being translated consistently and in line with the glossaries, and our IT team put any required fixes in place.
Craig and Sam stayed in regular contact throughout this process, which they describe as “a collaborative journey”. As Craig says:
“We’re pioneering and learning as we go! Rubric did a lot of trial and experimentation to find the best engine for each language. And they’re monitoring it with each release to see if anything can be improved.”
With the custom MT engines trained and ready for action, localizing a release is now a two-stage process. First, the priority UI and release notes are translated by humans. Then, the online help is translated by machine, drawing on those quality human translations.
For added peace of mind, we have robust quality checks in place. While the budget doesn’t stretch to a full human review of all MT content, automated checks flag up specific issues that might need human intervention (such as where the MT has suggested a different translation for a key term).
The result: Happy customers in half the time
Previously, a full release for NICE typically took five to six weeks with human translation. With Assured AI, that was cut in half: the human translation stage now takes around two weeks, with the MT content delivered shortly afterwards.
The faster turnarounds and reduced human effort have delivered significant cost savings too, enabling Craig to work within NICE’s budget. Rubric’s quality reports inform Craig of where MT translations are working well and where extra curation effort could be undertaken to improve quality overall. This is reassuring as it allows Craig to focus spend and effort where NICE’s end customers are focusing, enabling a targeted approach to quality translations.
“We’ve been very happy with the results. I feel much happier about this solution than any real-time translation tool we could have used.”
Craig Laurence
NICE Technical Publications Manager
Rubric can help ease your time and budget pressures with AI solutions you can trust. We partner with you to understand what’s most important to you and what level of quality you really need. Then, we tailor the perfect mix of AI and human input for translations that are on time, on budget and on brand.
“Rubric are THE champions of investing in partnering to deliver ideal solutions. The longevity and depth of our relationship make all the difference. It feels like a real partnership, and I know they’re always working in our best interests.”
To learn more about how Rubric can help you reap the benefits of AI localization, while mitigating the risks, visit our Assured AI page.