How AI is Reshaping Localization
5 Leading Tech Companies Share Practical Lessons for Building Hybrid Workflows
What does ‘AI in localization’ really look like? For many localization teams, it still feels like uncharted territory. Should you automate translation? How do you maintain quality? Where do human linguists still fit in all of this?
In this guide, five experienced leaders from Canva, Atlassian, Slack, Sinch, and Undertow open up about what’s actually working, what’s not, and what they’ve learned while integrating AI into multi-faceted workflows and international content strategies.
What You’ll learn:
-
How AI is already being used for translation, QA, routing, and content creation
-
What the biggest implementation challenges are, and how to overcome them
-
How human roles are evolving in the era of AI-assisted content
-
Practical, step-by-step advice for starting your own AI integration journey
This guide will show you how to turn AI into a strategic advantage, reduce bottlenecks, and build localization workflows that are scalable, efficient, and still led by human insight.
How Localization Leaders Use AI Today
Across all five companies interviewed, one thing is clear: AI isn’t here to replace you, it’s here to make your job easier.
It’s already reshaping how multilingual content is produced, reviewed, and scaled. But while adoption is widespread, the ‘how’ varies depending on team size, tech maturity, and internal goals.
AI in Translation Workflows
Most leaders in localization already have experience integrating AI into translation workflows with machine translation (MT), but what’s emerging now is a layered approach where MT is just one piece of a larger system.
To keep up with growing content demands without increasing headcount or budgets, teams are designing sophisticated workflows that blend machine translation, large language models (LLMs), and human post-editing.
At Atlassian, this approach is already in action. By combining AI-powered machine translation with targeted human post-editing, they’re able to manage high volumes of product content efficiently and cost-effectively, especially as their product suite and global user base continue to grow.
Canva takes this even further, using a multi-step AI workflow. There are two steps involved: one handles the initial translation, and another edits it. Originally, the pre-translation used neural MT for the base layer to get the most accurate output possible. However, later LLMs were brought in to refine and make the translation more fluent.
Automating QA, Bug Detection & Routing
AI is also transforming quality assurance (QA) and workflow management. The Slack team uses AI not just for content generation, but to pinpoint issues in source files, where they’re piloting bug resolution automation, as well as streamlining Linguistic Quality Assurance (LQA).
They've even embedded a prompt interface directly into their Translation Management System (TMS), letting teams generate English variants instantly or flag critical issues.
“The impact of leveraging AI in translation, and also AI in processes, is that we can go from localization taking X amount of days to taking a number of minutes.”
Anca Greve
Senior Director of Global Expansion
.png)
She stresses that automation isn’t about replacing people, but freeing them from repetitive tasks so they can focus on high-impact work, and quantify time back so it can be dedicated to more complex, human-heavy, relationship-management type work.
Content Scaling and Efficiency Gains
Sinch reports a staggering 300% increase in content output since adopting AI, all with the same team size. Now they are focused on customizing prompts and designing multi-path AI workflows to implement a true continuous localization strategy, increase performance, and give more flexibility to localization.
“AI is allowing our team at Sinch to implement a continuous localization workflow, increasing our performance, reducing waiting time to 0 while keeping quality, brand voice and style.”
Alfonso Gonzalez Bartolessis
Senior Localization Manager at Sinch
.png)
At Undertow, AI agents are now being trained to mirror brand tone and terminology, translating content more efficiently, and ensuring consistency while reducing manual input.
“AI agents can translate the new content in a style that is similar to what was done in the past and approved, using the right terminology for that brand.”
Nicola Calabrese
Founder & CEO
.png)
.png)
No One-Size-Fits-All: Why Custom Infrastructure Matters
There’s no plug-and-play AI solution for localization.
Every company we interviewed had to build or adapt their workflows to make AI actually deliver value. Off-the-shelf tools can help, but none of them support the full complexity of multilingual content operations, especially when it comes to integrating with translation memory, enforcing terminology, or aligning with brand voice.
"We’re using some out-of-the-box tools, but there’s definitely no standalone solution that covers everything from start to finish. We’ve had to build a lot ourselves.”
Michael Levot
Head of Localization at Canva
.png)
At Canva, the team experimented with an open-source agent framework (BrowserGPT) that let an LLM interact with their product user interface (UI). It got them about 90% of the way there, then they layered in their own localization-specific use cases.
Slack built a custom prompt interface directly into their TMS, giving internal teams access to AI tools that fit their exact workflows, whether that’s generating English variants or prioritizing QA bugs.
At Atlassian, the localization team is training its own LLMs using company-specific data and product UI, with the goal of producing higher-quality outputs over time and reducing reliance on generic AI behavior.
“The real opportunity lies in targeted training… the engines just need more data, training, and exposure to our unique products and product UI.”
Melanie Heighway
Head of Internationalization at Atlassian
.png)
The bottom line: if you want AI to work in localization, you can’t just plug it in. You need to train it, test it, and embed it into the systems and workflows your teams already use.
Real Challenges of AI in Localization
AI is already reshaping localization. But integrating it isn’t as simple as implementing a tool and watching efficiency skyrocket.
Behind every AI success story, there’s months of iteration, stakeholder education, and trial-and-error. While the upside is real, so are the roadblocks, especially for teams without technical support or buy-in from leadership.
Across all expert interviews, five challenges emerged again and again: quality, data, tools, people, and trust.
Human Oversight Remains Essential
Even the most advanced AI models need human review. Errors in tone, terminology, and meaning are still common, especially in customer-facing content.
AI performs well with widely spoken languages like Spanish, French, and German. It also shows promising results with high-resource but complex languages like Japanese and Korean. But for lower-resourced languages, particularly from regions like Africa or India, quality often drops significantly.
At Atlassian, AI-generated content is fully reviewed by linguists to maintain consistency and brand tone.
“Review time, at this time, is still double or triple what we would normally see of a human translator… largely due to the fact that there’s still limited training on our LLMs and data."
Melanie Heighway
Head of Internationalization at Atlassian
Tools & Models Aren’t Built for Localization
Even as AI advances, most tools on the market aren’t built with localization in mind. Standard LLMs can produce fluent content, but without context, they often miss the mark on tone, terminology, or functionality.
“If you don't provide the right terminology, you're gonna get generic translations.”
Nicola Calabrese
Founder & CEO
Translation memory, glossary enforcement, and visual context still matter. And while AI tools can mimic language patterns, they can’t automatically understand brand voice or product-specific nuance without proper training.
The challenge isn’t just getting the words right; it’s making sure your AI knows what those words mean for your product and audience.
“In localization, we don’t have a translation problem, we have a workflow management problem, and that’s what we really have to design for.”
Anca Greve
Senior Director of Global Expansion
AI Adoption Still Faces Internal Resistance
Rolling out AI in localization isn’t just about the technology; it’s about shifting mindsets.
Teams often face skepticism from leadership, reviewers, and even linguists. Some worry about job displacement, while others don’t see how AI supports business outcomes.
Experts recommend:
-
Starting with internal pilots to demonstrate value
-
Framing AI as a support tool, not a replacement
-
Sharing early wins to build credibility across teams
“Be ahead of the curve, push past that trepidation… experiment more rapidly and get to a good outcome before someone pushes you into a bad one.”
Michael Levot
Head of Localization at Canva
Scaling Quality Is the Hard Part
AI can handle volume, but maintaining quality as you scale is still an open problem. Without structure, speed becomes noise. But with the right guardrails, AI frees up your team to focus on what matters: cultural validation, campaign refinement, and testing.
“If repetitive tasks are taken care of by AI, you can focus on proper documentation and training for the team.”
Nicola Calabrese
Founder & CEO
The Future of AI + Human Linguists
The future of localization isn’t human or machine. It’s human with machine.
“We definitely want to assist, not replace people and processes using AI.”
Anca Greve
Senior Director of Global Expansion
As AI continues to evolve, so do the roles of the people who work alongside it. Across every expert we spoke to—from Canva and Slack to Sinch and Atlassian—the message is clear: AI is not here to replace human linguists, but it is changing the nature of their work.
Linguists are no longer just translators. They are becoming quality managers, cultural consultants, and workflow designers. In a world where AI can translate at speed, the value of human oversight is precision, empathy, and brand alignment.
Cultural Relevance Still Needs a Human Touch
Even as AI gets better at translation, it still can’t grasp nuance, emotion, or cultural specificity like a human can. Fluent output isn’t the same as effective communication, especially in customer-facing content.
That’s why teams at Atlassian, Canva, Slack, Sinch, and Undertow continue to rely on human linguists to ensure tone, context, and cultural alignment are spot-on.
“Linguists are not going anywhere anytime soon. Their expertise is 100% critical, especially for maintaining cultural relevance and handling nuanced content.”
Melanie Heighway
Head of Internationalization at Atlassian
At Canva, AI helps flag strings that need review, but linguists still make the final call. At Slack, editors handle culturally sensitive features. And at Sinch, teams route high-priority content directly to expert reviewers.
New Roles Are Emerging: Prompt Engineering, Workflow QA, AI Evaluation
As AI handles more of the repetitive work, the role of the linguist is evolving… from translator to strategist.
"Before we had teams made up of translators who focused on localization or transcreation. Their roles are moving more towards program management, prompt engineering, TM/TB management, design of quality strategies, and injecting high quality, validated and verified input into our systems."
Alfonso Gonzalez Bartolessis
Senior Localization Manager at Sinch
Linguists are now wearing new hats: prompt engineers, workflow QA leads, cultural consultants. They’re actively shaping how AI fits into localization workflows. Their ability to guide outputs, design QA processes, and create feedback loops makes them more essential than ever.
“Linguists are going to start being like quality leaders. They're gonna be consultants for cultural relevancy.”
Nicola Calabrese
Founder & CEO
Practical Advice for Getting Started
For many teams, the hardest part of adopting AI in localization isn’t technical; it’s knowing where to begin.
If you’re leading a small localization function or managing a team without in-house AI expertise, it’s easy to feel overwhelmed by the complexity, jargon, and hype. But every expert we spoke to emphasized the same thing: start small, stay practical, and focus on what will actually move the needle.
This section offers a step-by-step roadmap to help you begin integrating AI into your localization workflows, without burning out your team or your budget.
Start with Low-Risk Pilot Projects
The best way to build confidence with AI is to test it in areas where the stakes are low but the impact can be measured. Think internal documentation, help center articles, or placeholder content, anywhere speed matters more than perfection.
Avoid rolling out AI across your entire localization pipeline all at once. Use early pilots to identify gaps, refine your QA process, and build internal buy-in before scaling to high-visibility content.
“It’s absolutely necessary to test, test, test.”
Nicola Calabrese
Founder & CEO
.png)
.png)
Upskill Your Team on AI Fundamentals
AI integration isn't just about plugging in a tool; it’s about equipping your team to work with that tool effectively. That means training on prompt design, evaluating AI outputs, and knowing how to intervene when things go off track.
“Integrating AI into our workflows means we really need our LSPs onboard with that, and it requires ongoing training for our linguists to make sure they are comfortable with this new technology to work successfully.”
Melanie Heighway
Head of Internationalization at Atlassian
.png)
Even basic familiarity with prompt structure, terminology management, and AI behavior can make a big difference. If you’re not ready to train internally, consider working with a consultant to get your foundations right.
Target High-ROI Use Cases First
Focus your AI investment where it clearly supports your business goals. This often means high-volume, low-stakes content—like FAQs or product strings that change frequently—or automating repetitive internal tasks like context gathering or QA flagging.
Rather than trying to automate everything, prioritize use cases where AI can really deliver the most immediate value and free up your team for more strategic work.
"Internally, we’re automating a lot of repetitive tasks and reinvesting that time and budget into producing brand photography for specific regions. The work is more engaging for our team and more relevant for our users.”
Michael Levot
Head of Localization at Canva
.png)
Always Keep Humans in the Loop
AI may be fast, but it’s not foolproof, and as already mentioned, humans are an essential part of the whole automation process. Hallucinations, mistranslations, and tone mismatches are still common. That’s why every expert we spoke to emphasized the importance of a strong human review process, especially for customer-facing or brand-critical content.
Human input ensures that your localized content resonates in-market, reflects brand values, and meets quality expectations. Use AI to get the first draft, but let people make the final call.
“We continue to heavily rely on our editors, just because we have translated content in minutes, it’s still not ready to go out, particularly with marketing content.”
Anca Greve
Senior Director of Global Expansion
.png)
Define Clear Success Metrics
If you can’t measure it, you can’t improve it. Whether it’s turnaround time, human review hours, translation quality, or stakeholder satisfaction, set benchmarks early and track progress as you experiment with AI.
Use data to advocate for your efforts, get internal buy-in, and refine your workflows over time. AI success isn’t just about velocity, it’s about sustainable, measurable improvement.
“Invest in robust data tracking metrics to measure the impact of AI so you can also demonstrate your ROI later and continue the investment and development there.”
Melanie Heighway
Head of Internationalization at Atlassian
Conclusion
AI is already reshaping localization, but success depends on how teams apply it.
The companies leading the way aren’t just adopting new tools. They’re aligning AI with their workflows, training their teams, and using automation to reduce friction, not oversight. The most effective strategies start small, stay focused, and prioritize quality alongside speed. If you’re just getting started, begin with one clear use case. Test, measure, and refine as you go.
“Metrics, metrics, metrics, metrics, always refer to the data as much as possible.”
Anca Greve
Senior Director of Global Expansion
As every expert in this guide made clear, AI’s role in localization is only growing. The value of human input hasn’t disappeared, it’s evolved. Now is the time to rethink how your team works, not just how fast it can deliver.
Want help designing a multi-step localization workflow that uses AI effectively, without losing control of quality or consistency? Reach out to Undertow to get started.
“The possibilities with AI are infinite. We are just beginning to scratch the surface.”
Alfonso Gonzalez Bartolessis
Senior Localization Manager at Sinch
.png)
