Inside Vantage: AI-native localization for marketing

Full transcript

Welcome and introductions

Marta: Everyone, and thanks for attending this webinar today. I'm Marta, the senior product marketing manager here at Lokalise. And I'm super excited to host this session about Vantage, a new workspace inside Lokalise, built for marketing and long form content.

Marta: Here is our agenda today. We'll start by covering why marketing localization is still slow, how fluency has become table stakes, and why we shouldn't treat marketing content as code when it comes to localization.

Marta: Then we'll introduce Vantage and show you an end to end demo. And then at the end, we'll wrap up with a case study, and we'll have some time for Q and A at the end as well.

Marta: A couple of housekeeping items. You can use the chat to leave questions during the webinar, but we'll answer them live at the end.

Marta: So if we don't get to them, don't worry, we'll follow-up after the webinar with some emails. Also, this session is being recorded and we'll send you a recording via email after the session. So don't worry if you have to drop at any time.

Marta: Today, they're joining me here on this webinar. Lukasz from our product team, Fei from our design team, and Victor from our marketing team.

Why marketing localization is still slow

Marta: So let's begin.

Marta: We all know that AI has transformed most industries, and it has definitely made translations significantly cheaper by a thousand times. But it didn't make your campaigns launch any faster.

Marta: While AI reduces the transition cost, it doesn't reduce the localization complexity. We've got the same bottleneck that we had a few years ago. It's just that now it's more visible.

Marta: So, localization is still a bottleneck, and this is a problem. And why that is, is three reasons. First reason is that content teams don't own the process because the localization manager is driving it, not them.

Marta: They submit a request, they wait, then follow-up, then wait again. And this leads to markets getting skipped and missed deadlines. Second thing is that historically localization platforms were built for software keys.

Marta: So, they were built for developers. That means key value editors. That means command line workflows. That means API pipelines. And all of these works great for product strings, but it doesn't work the same way for a campaign brief or a landing page, blog posts.

Marta: And the third thing is that we are having more content and more markets. So volume is going up, market count is going up, and the localization step in the middle stays exactly the same size. So every campaign feels the drag.

Marta: And at Lokalise, we've worked alongside localization teams for years, since 2017.

Marta: And these three things haven't changed. What has changed is how much content is getting through them.

AI raised the stakes on discoverability

Marta: So AI hasn't just changed the cost side. It has raised the stakes on discoverability as well. Two figures that I want to share with you today that are from our latest white paper are that three out of four internet users are native English speakers, and that 70% of searches now end without a click, answer directly in the language they were asked in.

Marta: So, an answer engine replies in the language it was asked in, and pulls from the sources in that language. What that means is that if your page only exists in English, it's not in the running to be quoted at all, however good the page is. So you become invisible.

Marta: So there's a shift.

Marta: Before it used to be about ranking. Now it is about being eligible. I think markets used to be about growth bets, and now it's about keeping the traffic that you already have.

Marta: So, I think markets isn't an ambition anymore, it's the defense.

From fluent to resonant: finding Language-Market Fit

Marta: So, the bar has also moved higher for quality. AI has made multilingual content production fast and scalable, and quality is now high enough that fluency isn't a differentiator. It's expected. There is a difference between correct, fluent, and truly resonant content.

Marta: Let me walk you through the three of them.

Marta: So direct translations are technically accurate, but they miss nuance. Fluent content is that content that reads as native, but oftentimes fails to convert. And language market fit is the resonant content, which is that content that performs in the market, is the content that AI answers, like AEO prioritize, and is what drives measurable business outcomes. And this is the way that brands are nowadays winning locally.

Marta: So, I wanted to show you some examples of this so that you can understand a bit better.

Marta: The first one is about direct translations. So Siemens localized an app button, the back button into German, and apologies for any German speakers because I'm not a German speaker, but they got the body part back instead of and the correct... The word is correct, but it just has the word... The wrong meaning. That means that the content feels out of place, even though it's technically correct. So this would be a clear example of a direct translation. So that can be achieved nowadays with AI, but that doesn't mean that that's gonna resonate in the market. It's gonna look out of place.

Marta: So the next step in the ladder of quality would be fluency. So another example of fluency towards language market fit would be globalized customer, EF language, that uses the former Sie, also in German, to address parents versus the informal do for students. So it's the same language. It's just two different ways. It's the same market.

Marta: But because we're speaking to two different audiences, we make sure that that is resonating to the market.

Marta: And another example would be that in German as well, Life360, a customer from Lokalise moved deliberately the German formal to informal to feel closer to the families. So again, technically you would use the formal, but because in this market and towards this audience, this would work better, you deliberately make changes and make sure that even though one might read fluent, the other one is resonant. And the last example I wanted to show you is that sometimes the answer isn't doing more, but is doing less. So there was an early Pokemon dub that turned the Japanese rice ball into a jelly filled donuts.

Marta: And that wouldn't go great because today audiences would want the culture kept rather than replaced. And this is just a reminder that more localization isn't always better. The job is to meet the expectations, not to make everything hyper local. And the job is to make sure that you resonate with the market.

Marta: Okay. So now that we know that nowadays to win locally and belong in markets, need language market fit. How do you know what's working and how to reach that stage?

Marta: So we've got three dimensions that we use to measure this.

Marta: Resonance, intent, and impact. So for resonance, do we sound like a trusted local peer? For intent, does the content do the job it was created for? And for impact, does it move the business, AKA conversion, retention, revenue metrics?

Marta: So Fluency is telling you the content is correct and language market fit is actually telling you whether it's working or not.

An infrastructure problem, not a writing problem

Marta: And here's the real question.

Marta: Once you know what to aim for, how do you actually hit it consistently across every language and forever?

Marta: Historically, the answer was people. Skilled translators and regional marketers who carried all of this in their heads. What the brand should sound like, what terms to avoid, when to be formal versus casual, and how a joke lands in each market.

Marta: And then when it worked, it worked because of individual expertise, but not because of a system.

Marta: And the problem is that this doesn't scale. It's slow. It's expensive. It's nearly impossible to repeat the same quality market aftermarket once the content volume keeps climbing.

Marta: So here's the line that I want you to take away. The line that is in the slides. That fluency is a writing problem, so better prompts, better editors, better writers can solve it.

Marta: Versus language market fit that isn't that. It is an infrastructure problem, and that means turning context, orchestration, and judgment into something that runs on everything you publish, not just pieces a skilled person happened to review.

Marta: And that's what we're building.

Marta: And this brings us back to where we started. AI made translation cheap. Yes, we know that.

Marta: But it didn't make marketing content easy to localize. Just because marketing content isn't code, it is long form, as we addressed. It lives in a CMS, doesn't live in a repo, and it changes constantly. A campaign page gets rewritten three times before it even ships.

Marta: Any marketers in the room will know that. And people who own it are the marketers and content writers, not the engineers. And they shouldn't have to think like one just to get one page translated. This is exactly what customers here at Lokalise told us over and over.

Marta: They asked us to give them something marketing can actually use, a way of email threads and spreadsheets, localization that keeps up with the content calendar and that no one makes them wait for.

Marta: And that's why we built marketing its own workspace. And that's Vantage. That's the workspace we built for marketeers and for long form content. So now I'm gonna pass it over to Fei so that she can walk you through all of this. Thank you very much.

Introducing Vantage

Fei Huang: All right. Thank you, Marta. Hey, everyone. I'm Fei, one of the product designers on Vantage. A new workspace inside Lokalise built specifically for marketing and long form content. So before Lukasz shows you the demo later, I want to give you a quick tour of what you're about to see.

Fei Huang: Lokalise has always been powerful for product, engineering, and localization teams. And what Vantage does is that it extends Lokalise to other content coming from marketing teams, legal, HR support, etcetera.

Fei Huang: So it's the same platform and the same language assets underneath, but a workspace designed around how everyone works. But here's a bit more context on where Vantage actually sits inside Lokalise. If you're already a Lokalise customer, know the classic editor. It's built for software strings, developer workflows.

Fei Huang: It has GitHub. It has Figma. That's expert. On the other hand, Vantage is the second workspace.

Fei Huang: It's, again with the same platform underneath with shared translation memory, security, billing, user roles. But it's purposefully built for long form and marketing content.

Fei Huang: I will dive a bit deeper into why Vantage. We have five specific topics, and I'll walk through each and every one of them. Let's start with built for marketing tips.

Fei Huang: First, you don't need to be technical to use Vantage.

Fei Huang: You don't need any training. You don't need to create tickets for anyone. Whoever owns the content can request and track their own initiative directly inside Vantage.

Fei Huang: If you do need to preview the content before it's exported, then we have in context previews where translations are ready, and you can actually look at the translation in the actual layout. This way, you can catch length problems, layout breaks, copy that doesn't fit in the design.

Fei Huang: On top of that, we go to where your content lives. We have integrations for Google Docs, Webflow, Contentful, etcetera. So you don't need to export anything from platforms, and you can just get them directly into Lokalise.

Fei Huang: If you do work with files, then we also have support for various file types such as Word documents, PowerPoint, HTML, markdown, and so on.

Fei Huang: Lokalise AI quality and review.

Fei Huang: Vantage leverages localizes language intelligence by translating the document in batches of segments. So this way, we can give AI the surrounding context it needs for better quality. Then what happens is that it will score each of the segments and route only the ones that fall short for review.

Fei Huang: So this means that actually most of the content will go through automatically, and reviewers can then focus on the edge cases where they... Where it needs their attention.

Fei Huang: We have shared assets across Vantage and Expert. So marketing engineering, they will draw on the same translation memory and approve translations.

Fei Huang: So whenever new translations are approved, it will add to the translation memory that will be shared between marketing and, you know, product and engineering teams moving forward. This way, your campaigns will always sound like the product and vice versa.

Fei Huang: Lastly, we have control that scales with you. So permissions, quality rules, and review requirements are set when you create a project. So whenever you add new content to set project, the same settings will apply.

Fei Huang: Adding to that, every action will be captured in the audit log, so you can always go back and review to check if you need to check anything, understand what happened, and so on and so forth.

Fei Huang: Alright. That's pretty much it. That is the workspace.

Live demo: Vantage end to end

Lukasz: Now let's see it in action. I will hand the mic over to Lukasz.

Lukasz: Thank you, Fei. And hello, everyone. My name is Lukasz. I'm a staff product manager here at Lokalise.

Lukasz: Let's jump straight in and see how Vantage works in action. Imagine I'm localization manager at a company called BIMA. I've just received an urgent request from our marketing department to translate a blog post into three target languages.

Lukasz: They need everything to be ready by end of the week. Let me just share my screen now and show you the blog post that requires translation. This is what I received from my marketing team.

Lukasz: I will now share our Vantage workspace. And this is where you manage your long form and marketing content localization.

Lukasz: At any point of time, you can always go back to expert, but let's remain in Vantage for the next few minutes. I'll create a new project. A project in Vantage is a container of your localization settings for a particular type of content. In my case, I'll be using it to manage translations for company blog posts.

Lukasz: The first step of project creation requires me to choose the path for the content source. I can choose between files, and there's multiple files we support. HTML, docs, markdown files, IDML, Google Docs integration.

Lukasz: Or we can choose the path of connecting to existing integrations where we store content. The choices are between content management systems, such as Contentful, Webflow, or Strapi, or an email automation tools like Braze or SendGrid.

Lukasz: Let's go ahead with uploading a file option.

Lukasz: I'll give this project a name, and let's just call it blog posts articles.

Lukasz: Our company's source language is English — specifically British English. So I'll define it here. I want to be very specific.

Lukasz: And we need translations into French, Spanish, and German.

Lukasz: Now let's go ahead to the next step.

Lukasz: This is where I need to upload the file I received from my marketing department.

Lukasz: Quick drag to upload the file, And we're moving to the next step. The next step is optional, but it's very important for context for AI translations. I have two options here.

Lukasz: I can either create or upload as an... A CSV with my glossary. As you can see on the screen, I already have done this step before, and I have 24 existing terms.

Lukasz: I can always update the terms from this step.

Lukasz: I can change the term, change the description, mark it as non-translatable, or provide a specific translation I want for a specific target language, or mark it as case sensitive.

Lukasz: Let's leave it aside for now. Another option is to connect translation memory.

Lukasz: So I can add a translation memory or connect it with the one I'm already using for the expert workspace.

Lukasz: So this is an important step. But for the sake of the demo, let's just continue right now. And last but not least, I'm asked to choose the translation process. I have two options available.

Lukasz: Either instant pro AI translations, where speed matters. We rely on AI translations and context that was shared before. Or we have an option of AI translation with human review. I want the latter because my marketing team always requires that a human looks at translations before they get published. For human review translations, I need to define the review scope.

Lukasz: We have four options here. We have the full review, where every single segment of the content piece is reviewed by an assigned contributor. We have an extensive review, where our proprietary AI scoring flags issues that are minor, major, or critical, and the reviewer is asked to review only those segments. We have a focus review where only major and critical issues are flagged for review, or essential, where reviewer need to care only about critical items.

Lukasz: I'll choose a focus review.

Lukasz: I'll set the deadline to be Friday.

Lukasz: That seems to get accomplished by end of the week. And I'll assign my reviewers to make sure that the right person will be a contributor to review translations.

Lukasz: For French, that's Jonas. For German, that's Thomas. And for Spanish, I prefer Laura.

Lukasz: And just like that, I'm triggering right now the AI translation process.

Lukasz: Okay. Now we have AI translations done, and I immediately see that 92% of all segments are approved by AI.

Lukasz: As an orchestrator of this blog post translation, I like to skim through source and target translations to just search whether I spot any obvious changes before I involve my reviewer. I speak Spanish, so I'll be focusing on the Spanish translations only.

Lukasz: On the left side of the screen, we're seeing the editor with the content units split into segments for the ease of review. On the right side of the screen, we're seeing the interactive context preview, where, basically, I'm seeing the HTML file rendered so that I can follow what's happening what's happening in the editor on the live preview.

Lukasz: I like to use the compare mode where I basically review side by side the source copy versus target translations.

Lukasz: For example, I don't like these sentences that long, so I'll try to act upon it.

Lukasz: I'll edit this translation by using this edit option from the compare mode, and I'll just shorten this translation.

Lukasz: I'll also make sure that I see the changes in real time.

Lukasz: Let's use AI to help me choose the more appropriate shortened version.

Lukasz: And let's say that I like the option two the most.

Lukasz: I check this, and now I approve the file. And here we go. A shortened translation directly in the interactive in context preview.

Lukasz: Now let's look at all the issues for Spanish target language. I see four of them.

Lukasz: Let's review the first one.

Lukasz: I see there's one issue, and I don't necessarily like how this translation sounds. So I'm gonna rephrase it.

Lukasz: I'm gonna choose the option two.

Lukasz: Here we go.

Lukasz: This is just a snapshot of how our editor experience works for the reviewer.

Lukasz: Now let's switch seats for a moment. And let's imagine I'm Laura, the Spanish reviewer that I assigned Spanish translations to have a look at. I received an email with the task for Spanish translation review with a total scope of 47 base words. And the project called blocks bot... Blocks bots articles.

Lukasz: I click on view task details, and I'm seeing a Vantage workspace that is basically an overview of all the tasks that are assigned to me.

Lukasz: That's the task that I just received into my inbox, but I also see there is another overdue task by twenty three days. Let's ignore this one for a moment, and let's focus on the one that I just created for myself, with the due date being Friday, end of the day.

Lukasz: Let's go to the editor. I'm welcomed with a very brief journey to help me understand how I use and take the most advantage of vantage. And then I'll jump straight into reviewing assigned content.

Lukasz: As you can see, I already have four segments pre filtered because these are the only four that were scored by our proprietary AI scoring engine with a major issue.

Lukasz: One is already approved because me wearing a previous hat of a localization manager, I already approved this translation. But as Laura, I can still see it, and I can unapprove or approve it.

Lukasz: And let's go to the next one.

Lukasz: Let's say I don't really like this transition, and I apply for a change here.

Lukasz: I like this option much better. And then let's just assume that I don't have any issues with translations for the last two pending segments. So I just approve the translations.

Lukasz: And just like that, I have my task closed. I can move on to the next task right now, which is the pending task we've seen on the screen before, or I can just finish the session.

Lukasz: And that wraps up a complete vantage end to end journey. I will now pass it on to Victor. Thank you.

Case study: localizing lokalise.com

Victor Tejeda: Lucasz and Fei have just shown you what the Vantage workflow looks like for one piece of content. What I want to do now is show you what happened when we applied that workflow to our own website at a larger scale. I'm Victor, and I work on SEO at Lokalise. Part of my role is making sure that lokalise.com can reach and serve audiences outside of our English speaking markets. For these projects, we selected 33 priority pages from lokalise.com and prepared them for German, Spanish, and French.

Victor Tejeda: That gave us useful real world tests. This wasn't a simple document. It was our website, our CMS, our terminology, our content that ultimately will represent our brand in public.

Victor Tejeda: That difference matters because localization normally becomes more complicated when you touch a production website. You are no longer dealing with words alone. You have structured CMS fields, navigation, reusable components, page layouts, review deadlines, and publishing checks. So we wanted to answer a practical question.

Victor Tejeda: Could we build a repeatable workflow around Vantage that our content and SEO team could actually keep using over time?

Victor Tejeda: Our starting position will probably sound very familiar to many marketing teams. We had localized products and plenty of localization expertise inside the company, but our website content was still available only in English. We also didn't have a s... Existing website localization operation that we could simply move into Vantage. There was no urgency, workflow to migrate, no established handoff process, and no assisting tooling connecting website content to translation back again. In one sense, that made the project harder because we had to define the process while running it.

Victor Tejeda: But it also made the test cleaner. We weren't trying to make an old workflow slightly faster. We were asking what the workflow should look like if we design it today.

Victor Tejeda: We started with a control scope, 33 priority URLs containing 60,000 English words going into three target locales, German, Spanish, and French. We didn't start by translating the whole website. That would have created volume without necessarily teaching us more. Instead, the initial phase was designed to test the operating model, how content enters Vantage, how it gets divided into manageable review batches, how the reviewers work with it, and how it returns back to our CMS, and what still needs to happen before publication.

Victor Tejeda: It's also important to be clear about what this project proves and what it doesn't. The immediate question was whether we could move a minimum amount of real website content through a controlled and repeatable process. And the long...

Victor Tejeda: Longer term goal was not just to complete these 33 pages. It is to make the next page and the next market easier to add. So it's an improving process.

Victor Tejeda: That is the real difference between a localization project and a localization system. A project finished when the files are delivered, but a system become part of how the website operates. This is the workflow we use. It is simple. Strapi, Vantage, human review, and back to Strapi.

Victor Tejeda: Strapi remains our source of truth. Our English content is written, updated, and structured there. Here is where many teams create a separate content operation where people have to copy text into spreadsheets or maintain disconnected documents. That's exactly what we solve.

Victor Tejeda: The Vantage integration in the second part pulls that content from Strapi into the localization workspace. Vantage then processes the page as a structured content, applies our existing language assets, including translation memory and review glossary, and creates target language content. Before beginning the batches, we spend time reviewing the glossary. That may sounds like rather preparation than delivery, but it is an important part of the workflow.

Victor Tejeda: AI can produce quickly a good output. The challenge is ensuring that product names, technical terms, and repeatable context are handled consistently across dozens of pages. The glossary and translation memory provide that source context...

Victor Tejeda: Shared context.

Victor Tejeda: Once the translations were ready, we were working with Argos, our language services partners for human review. We divided the project into seven batches. Each batch was kept to a manageable size, generally around 10,000 words. That gave the reviewers a predictable unit of work and allow us to track progress without treating 60,000 words as one enormous delivery.

Victor Tejeda: Our human review target was five working days per batch. And I want to be precise about the number. Five days was the review turnaround, not end to end promise that every page will be published in five days after creation. Of course, that can be speed up depending on the language service provider and your own limitations. Translations itself can happen extremely quickly. Human review, the final CMS checks, the page previews, and publishing still have to be planned.

Victor Tejeda: After the review, the translated content was exported back to Strapi as drafts.

Victor Tejeda: That draft step is important. We did not want translations to go public automatically. Returning them as drafts gave us a final checkpoint to preview the pages in context, verify links, components, the metadata, layout, and the overall page experience. We also learned several practical lessons. First, validate the CMS content model before starting. We encountered cases where an export failed because a field in the English source content contained a value that the Strapi schema did not accept. Once the source field was corrected, the export worked.

Victor Tejeda: That was not a translation quality problem. It was a content model problem.

Victor Tejeda: Second, Agree on batching rules before sending content to review. Seven smaller batches were easier to manage than one large delivery, But batch ponders and deadlines need to be clear for everyone on both. They can be speed up for sure. Third, reserve time for in context QA.

Victor Tejeda: A translation can be correct and you still need adjustments once you see it inside the page. German copy may take more space. A heading may wrap differently or or a component may behave weird. So although this diagram looks like a straight line, there is a quality checkpoint before publication.

Victor Tejeda: The main advantage is that the content begins and ends in the same CMS.

Victor Tejeda: Vantage manages the localization work in between without forcing our marketing team to create parallel content system. Here is the operating picture behind the project. These numbers are useful, but the most interesting result is not simply the numbers of words translated. 60,000 words across three locals represent a substantial amount of target language content. The belly of Vantage was that we didn't have to manually coordinate every every individual page and language combination. The work will move through one repeatable path using the same terminology and the same review structure.

Victor Tejeda: The five day review cycle also needs the right context. Saying that AI can translate quickly is no longer especially surprising. The difficult part is everything around the translation, preparing the content, applying the right language context, assigning the review, following the process, and returning the results to the CMS, then checking it before the publication. The stronger and more accurate result is this. We move from an undefined website localization process to a predictable review cycle for batches up to 10 k words.

Victor Tejeda: That predictability is what makes planning possible.

Victor Tejeda: There are also things we will do differently next time. We will validate short fields and CMS schema requirements before launching the first batch. We will finalize the glossary earlier. We will define batch sizes and reviewer availability before importing the full scope. And we will include final in context QA explicitly in the project timeline, rather than thinking about it as the last small step.

Victor Tejeda: Those lessons are part of why we run this project internally.

Victor Tejeda: Dogfooding isn't about presenting a frictionless success story. It give us direct experience of where the product works and where the overall process still needs improvements.

Victor Tejeda: The result is a workflow we can repeat as we add pages, update existing content, or expand into another market. And that connects back to the main idea of this session. Localizations will not be a separate project that marketing has to restart every time you launch something.

Victor Tejeda: The goal is for localization to become a property of the content workflow itself. Content is starting the CMS, moves through translation and review, and comebacks ready for final q and a... QA and publication.

Victor Tejeda: That is what this first phase demonstrated for lokalise.com. And with that, I will hand it back to Marta.

Customer proof: Headspace

Marta: Thank you very much, Victor. So, before we finish, we just wanted to share a quote from one of our early customers. Just to show you that this isn't just on our own website, but we have customers that are already live on Vantage too. And Riz Garcia, a production coordinator at Headspace, tested the portal on Vantage early on. And she was sharing with us that it was pretty straightforward, very useful for marketing, and everything works so smoothly.

Marta: And something that's worth pulling out is that she didn't need onboarding to use it. And she was saying, I don't have any notes. And this is coming from someone whose actual job is catching problems. So this is exactly the outcome that we were aiming for.

The results, and what's next

Marta: So let's put some numbers so that you can see that the localized standard actually translates into Vantage as well.

Marta: For speed to market, we are going from four weeks to under two days to launch a new market. 92% plus is the standard for acceptance rates of what Vantage produces without rework.

Marta: Cost is also roughly 80% savings versus human translation and up to 80% less post editing. So those are great numbers behind. So let's wrap up.

Marta: Fluency is table stakes. Language market fit is the bar. That means the content that does the job in the market, not just the content that reads well.

Marta: Getting there is an infrastructure problem, not a writing problem, as we've seen.

Marta: So context as managed assets, rules that root by risk, and people on the judgment call is what is gonna make you win in local markets.

Marta: And Vantage is that infrastructure that gets you to language market fit.

Marta: So a workspace in Lokalise that your marketing team can run without training.

Marta: So what now? What can you actually do after this webinar? We say get started with Vantage.

Marta: So, if you're already a Lokalise customer, Vantage is going to be enabled by default on your current plan. You can open it from the workspace switcher and take the guided tour. No procurement, no new contract. You can get started right away.

Marta: Second, you can pick one content type, not all of them. Pick a campaign landing page, a one pager, blog article, and set the workflow and review scope once. And then you run the next one through the same setup.

Marta: And thirdly, for those engineers and developers in the call, bring your marketing team in. Give them the self serve portal directly so that localization stops arriving as a ticket and it starts being something that they just do.

Marta: And now it's time for our q and a. We're going to answer all the questions live and thank you for watching.