A content localization strategy for global training has six components: a decision on which languages to start with, a roadmap tied to the business calendar, a team structure that separates content creation from translation management, a pilot process before full rollout, a review system that combines language competency with industry knowledge, and a technology layer that reduces manual effort without removing human judgment.
Localization means adapting content so it makes sense to the person receiving it, in the context they are working in. Translation moves words from one language to another. Localization moves meaning. Those are different problems, and organizations that confuse them end up paying for both. This article covers each component in order, drawing on how Keune Haircosmetics scaled to 23 languages across 70 countries, and what Derek Bruce learned building global programs at ABN Amro and Tesco.
Note: This article was published by Easygenerator. EasyTranslate, mentioned in the dedicated product section below, is an Easygenerator feature.
Localization and translation are not the same thing, and the difference is expensive
Organizations that treat localization as a translation task spend money twice: once to translate, and once to fix what the translation got wrong. That is the most common and most avoidable cost in a multilingual e-learning program. We covered this in a recent webinar.
Derek Bruce, Chief Learning Officer at Easygenerator, described the default he observed early in his career at ABN Amro, a Dutch company with 17 international locations. The approach was straightforward: the head office determined the content, finalized it, and then sent it for translation. As Derek put it, “It was very much, almost like the head office kind of determines the content and you just translate rather than localize, which are two very different things.”
The practical result of that approach is content that is technically accurate in another language but contextually wrong. Industry-specific terminology gets translated literally rather than adapted to local usage. Cultural references that make sense in one region mean nothing in another. Measurement units stay imperial when the receiving country uses metric. Examples reference company names, practices, or events that the local audience has never heard of.
The organizational cost goes further than relevance. When learners receive content that was clearly designed somewhere else, for someone else, it sends a message about how much they matter. Derek described the effect directly:
In global organizations, that signal affects engagement with learning and with the organization overall.
Jon Withrington, Global Customer and Capability Education Manager at Keune Haircosmetics, shared the same experience in the same webinar. When Keune sent content to an external agency for translation, the output was linguistically correct but lacked the precision that comes from knowing the industry. “When the translations came back, those translations lacked the precision because, of course, we go back to the word, like lingo, you know, these local nuggets.” The solution was not a better translation agency. It was assigning a person from the local education team in each country, someone who understood both the language and the industry, to review and adapt the content.
How many languages to start with, and which ones
Start with three or four languages, not the full list. Trying to launch in ten or twelve languages at once is the most common early mistake, and the one that makes the entire process collapse before it produces any value.
Jon learned this at Keune. The platform now serves 70 countries in 23 languages, but it did not start that way. “I know in the beginning we started off with English, German, French, and Spanish. And then the year after we introduced another two or three more languages. So don’t try and do everything at once. Really have a roadmap in place and think strategically about this.”
The selection criteria are straightforward. Start with the languages where your audience is largest, where your business impact is highest, or where you have local team members who can support the review process. All three of those criteria can point in different directions, so it helps to pick based on a combination of reach and reviewability.
Worth remembering
A language you can deploy and review well is more valuable than a language where you cannot manage the quality of the output.
Six to eight months is a realistic timeline to get a single new language properly set up in a localization process, from first draft through pilot to full deployment. Jon’s estimate at Keune was that range for each language, from point to point. That timeline matters for planning: if you are trying to add four languages in the first year, the work cannot be done sequentially, which means you need either more resources or a longer timeline.
Derek’s framing for prioritization is practical: “Work out the biggest headcount, work out really where is the biggest impact. Involve those people in it as well. And once you have got that and you know things work, then you can start evolving it as well.”
How to structure your localization team
A working localization team has two distinct functions that should not be handled by the same people. Content creation and translation management require different skills, different timelines, and different relationships with local teams.
At Keune, Jon’s team is divided along exactly that line. The customer and capability team creates the content and works with internal stakeholders to get the product information, regulatory requirements, and educational objectives right. The digital team owns the translation management: they run the project from start to finish, manage the timeline, meet with translators at least once per week to keep things on track, and make sure local teams have everything they need to do their work.
“What our digital team does extremely well is ensuring that the teams globally, our education teams and translation teams, have all the tools and support that they need. So they are set up for success.”
That weekly cadence with translators matters more than it sounds. With 23 languages in play, things fall off the radar if someone is not actively managing the pipeline. A missed check-in becomes a delayed translation, which becomes a delayed launch, which means learners in that market wait longer for content that was ready in other languages weeks earlier.
Derek added a second dimension to team structure that is less about headcount and more about mindset. L&D teams working from a central office can fall into a pattern where they treat their own output as the standard and local teams as reviewers who are there to approve rather than contribute. He described seeing this in practice: a “we’re better than” mindset that makes collaboration harder without anyone intending it. The fix is treating local team members as co-creators from the design stage, not just reviewers at the end.
Derek also described a practical tool that helps regardless of team size: a design blueprint that sets the defaults for look and feel, tone of voice, and how content should come across. That blueprint gives anyone contributing to localization a reference point so that individual judgment calls stay within a shared framework.
How to pilot before you scale
Piloting with a small group of countries before full deployment catches the problems that no amount of central review will find. It also builds the confidence in the process that makes scaling faster, because you are not troubleshooting basic issues in twelve languages at once.
Keune’s pilot structure covers different regions deliberately. Jon’s team tests with North America, the UK, Spain, and Brazil before rolling out globally. Those four markets cover different languages, different regulatory environments, different cultural contexts, and different relationships with the Keune brand. If content works across all four, it is much more likely to work everywhere else.
The pilot is also where the timing expectations get tested. Jon described being very deliberate about this: weekly standups during the pilot phase, quarterly reviews of the full cycle, and ongoing communication as a collective rather than one-off check-ins. That structure exists because localization problems tend to surface during the pilot if you are paying attention, and tend to surface at launch if you are not.
Derek flagged a timing issue that undermines most pilots: giving local teams a week to review near-finished content is not a real pilot. “One of the worst things is to go: here’s our 99.9% finished content, you’ve got a week to make any changes. And they’re like, well hang on a second, this doesn’t make sense.” A real review requires time built into the process before the content is treated as done. That means building review cycles into the project timeline as first-class activities, not squeezing them in at the end.
His experience at Tesco gives a sense of what serious piloting looks like. For a leadership development program covering 3,000 leaders across the UK, Slovakia, Slovenia, Poland, and India, Derek and his team ran three or four pilot sessions in different locations before the final product was signed off. Each session involved participants from the relevant locales, and each session produced changes. The content that launched was materially different from the content that went into the first pilot.
How to build a review system that catches local nuances
Worth remembering
Build your review system around people who understand both the language and the industry, not just one of the two.
Content that is technically correct but contextually wrong is the most common failure point in multilingual training, and no central team or translation tool can catch it without local input.
Derek described a single word that illustrates why. “Banter” is a common term in UK workplaces, used to describe friendly teasing between colleagues. It is entirely unremarkable in that context. Outside the UK, it carries a different meaning, or no meaning at all. When Derek’s team was building an inclusion program at Tesco that referenced banter in the context of workplace behavior, the word only got flagged when people from other locales joined the review process. “By having the different locales involved in the review and design process, it was evident that that just doesn’t make sense. How do you get the message across without using that word but still have it make sense?”
The fix is not just adding a local reviewer to the end of the process. It is involving local team members early enough that they can shape the content before it gets finalized, rather than just approving or rejecting something that was written without them.
Jon’s review structure at Keune makes this concrete. Each country has an assigned person from the local education department who is responsible for translation review. That person is not simply bilingual. They are someone who works in the industry, knows the product terminology, and understands the local professional context. “For us the key was also working with our local education teams. And in each country we have an assigned person from the education department who is doing the translation, sometimes if their competency is high enough, or in conjunction with a translator. But they make sure that everything is adjusted and adapted as needed to the local context.”
That structure takes time to build. Jon acknowledged that the first year was largely trial and error, that in the early days one or two people were building everything and piloting was not robust. Getting to a review system that actually catches local nuances required building relationships with local teams, establishing clear roles, and creating enough trust that local reviewers felt their input was genuinely valued rather than treated as a formality.
Derek described the same dynamic as collaboration rather than training: “I don’t know if the word train is the right word, if it’s in the sense of the design outcome. Maybe it’s collaborate with or actually work with. Because train might imply that we get them to do what we want to do anyway, when in fact sometimes it may not work.”
Where AI fits in a localization workflow, and where it does not
AI speeds up the translation layer significantly, but it cannot replace the judgment of someone who knows the local context. Both speakers landed on the same conclusion from different directions, and the cost data from Keune gives the most concrete picture of what the efficiency gain actually looks like.
Before Keune adopted an AI video creation tool for their content production, a single production cycle cost between €20,000 and €30,000 with a usable lifespan of two to three years before the content needed to be redone. The investment they made in AI changed that substantially. As Jon shared in the webinar: “We spent in total €24,000, we produced over 800 interactive videos and it saved us around €800,000.” The efficiency came from being able to produce more content, to update it faster, and to localize it across languages without rebuilding from scratch each time.
The caveat that follows directly from that result is the one both speakers were clear about. AI output is a starting point, not a finished product. Jon described the risk of treating it otherwise: “Everything that you do with AI needs to have a human touch. Because at the end of the day, we are in the business of curating brain changes.” For Keune specifically, the review has to be done by someone who both understands the language and works within the hairdressing industry. A general translation tool handles neither the professional terminology nor the brand voice correctly.
Derek described an unexpected benefit of AI in multilingual video that his team discovered during a production review. When they watched back a multilingual video, they noticed that certain AI-generated voices were being used in patterns that felt subtly biased: some voices were consistently paired with positive framing, others with negative. With a human voice recording, fixing that would have meant going back to a studio. With AI, it meant reassigning voices and regenerating. “It just meant literally going, okay, that voiceover can now be this. And it was done, and that was within a couple of hours.” That kind of rapid correction is only possible when the production is not dependent on scheduling a voice artist.
The overall framing Derek offered is the right one for any team introducing AI to a localization workflow: “AI is a tool. It’s not the answer all the time. So use it wisely.”
How to manage stakeholder expectations when AI enters the workflow
The biggest challenge when adopting AI for localization is not the technology. It is the assumption that AI makes everything instant.
Jon described the stakeholder dynamic at Keune when he was introducing an AI video creation tool. Getting buy-in required showing the numbers and being clear about what the tool would and would not do.
The questions that came up went well beyond the output quality: data compliance, how the voice library worked legally, how real the avatars should look, whether using an avatar meant a human actor had lost a job.
The Harry Potter wand problem, as Jon called it, is that people who are not close to the production process often assume AI delivers a finished result with minimal human effort. “AI is really like the Wand of Harry Potter, you know, you just whoop, and you get the result and it’s ready to go.” Setting expectations correctly before the first deliverable arrives is much easier than managing them after a stakeholder has seen a prototype and started requesting changes that are not technically possible without significant rework.
Jon described a pilot review session where a stakeholder kept requesting changes to how the avatar moved and spoke, getting more specific and more demanding with each round. “I got to the point I really had to sit down with this individual and say: with respect, it’s now taking so long to try and do these impossible things. Maybe it’s actually easier to hire a production company and just do this externally, which actually defeats the whole point of using AI.” That kind of conversation is easier to avoid than to recover from.
The practical approach is to show the numbers first. Jon’s advice for winning stakeholder buy-in is direct: “Show them a beautiful video, tell them what they need to invest, tell them what they think they can get in return. As long as you have the numbers, everyone’s happy.” Once the ROI case is made and the timeline expectations are set clearly, the conversation about what AI can and cannot do becomes much more manageable.
EasyTranslate: translation that stays inside your authoring workflow
EasyTranslate is Easygenerator’s built-in translation feature. It translates courses into 75 languages without requiring any content to leave the authoring environment. The practical difference between this and sending content to an external translation tool or agency is that reviewers work inside the course itself rather than in a separate document. They see the content in context, make targeted edits where the translation needs adjustment, and the updated version is ready to publish without any reimporting or reformatting.
When content changes, whether a product update, a process change, or a regulatory requirement, you push the update in the source language and translate it again. Dynamic SCORM means small edits go live in the LMS without requiring a full re-upload of every language version. That removes the maintenance bottleneck that makes most multilingual programs expensive to keep current.
For teams building a review workflow with local subject-matter experts, EasyTranslate gives reviewers access to the course directly rather than routing feedback through email threads or document comments. The process that Jon described at Keune, where a local education team member reviews and adapts translations, is exactly what the tool is designed to support.
