Entering a Market Isn't the Same as belonging in It

Full transcript

Introduction

Sophie: Localization has always been a growth driver. But until recently, it was simply too slow and too expensive to do everywhere. And that meant brands settled for good enough and in a few markets. In the last two years, generative AI has radically pivoted the economics of localization and the opportunity for brands. 

Let me step back. I have spent my career building international companies pushing for growth, and localization was always an operational tax. It took weeks to open a market, it would cost millions of dollars to support 10 languages, content sitting in queues, team chasing marketing waiting on engineering. So most companies translated a little into a few markets and accepted that good enough would have to do.

Generative AI removed the compromise. 

Sophie: It has radically opened the world of possible: the operational swiftness, the economics and also the bar for consumers' expectations. Companies working with us opened a new market in under two days, not four weeks. They save around 80% compared to the cost of human only translation. And critically, quality went up, not down. 

With the right setup, more than 90% of translations are accepted without human intervention. But higher efficiency is only one of the benefits. What modern localization actually buys, what businesses actually want, is reach, growth, content that customers engage with, which makes brands more visible, which converts better, content that brings 40% more revenue from those markets, revenue from places you previously could not serve well. Today, 3,000 companies run on our platform, including 10% of the global Fortune 500. 

Localization has stopped being an operational tax you manage. 

Sophie: It is now one of the highest return growth decisions a leadership team can make, and boards are starting to see it that way. Ambitious brands, they're woken up to that. They're seizing the opportunity, and by doing so, they're raising the bar for everyone to compete. 

Recently, I sat with two of our customers: a global financial services brand with a strong presence globally and a fast growth software company. Both have solid international business, yet both said their success will require them to move from a well recognized global brand to becoming local and authentic in each market and for each audience. The competitive advantage will move from your content being fluent in a market to whether you look like you belong there. We call it language market fit. 

Language-Market Fit

Sophie: That's when your brand truly belongs in its market. Companies today can be found in three categories. 

The first one is direct translation. Words are accurate, the meaning survives. And this is what companies have been buying for twenty years. And at real scale, it is genuinely difficult and it still reads as foreign. So your customer understands it but knows that it was not written for them. 

The second one is fluency. It reads naturally. It uses the words people in that market actually use and nothing about it feels imported. This is real progress and it's where AI has moved the entire industry in about two years, which is exactly why it's no longer a competitive advantage. Your competitors are already at this level or they will be inside a year. Fluency has become the floor, not the ceiling. 

The third category is language market fit. Here, the message itself is adapted to the audience, to the market, to the channel and to what you're trying to achieve commercially. Because the argument that persuades a buyer in Tokyo is not the argument that persuades one in Paris. At this level, you're not translating your message, you're making it land. 

And there's a new reason this matters more than it used to. AI answer engines prioritize sources and the language of the question with relevant local content. If your content does not read as though it belongs in a market, you're much less likely to be found there at all. 

Fluency used to be about how you sound. Now, it decides whether you're seen. Most companies today are somewhere between the first stage and the second. The third is still an open opportunity, and this is what we're building at scale. This is what Sahil is going to show you. 

Lokalise's evolution

Sahil: I'm Sahil. I'm the chief product officer at Lokalise. Lokalise is an AI native localization platform trusted by companies like IBM, JPMorgan, Navan, and Life360. When we speak to a business with tens of thousands of employees operating in 20 markets, they understand the value of language market fit. But they don't know how to manage the complexity behind it. 

Organizations like these need us to connect to everything they already run, code repositories, design tools, content platforms, support systems. So we offer hundreds of integrations, a best in class API, and an automation to run localization continuously alongside every release instead of in a separate sprint. 

They need governance, one source of truth, clear control of who can do what in which market. Every change is recorded and every result is measured. They also need proof of reliability, security and compliance first, SOC two, ISO 27,001, GDPR, zero data retention, granular permissions, and a full audit trail. Your data is never used to train shared models. 

Lokalise can meet these requirements because engineering teams demanded them of us long before anyone else did. Lokalise started in 2017 as an innovative solution for engineering teams. Developers chose it because it was precise, it was fast, it was continuous, and it fits smartly in the way they already worked. 

From there, we brought design stage localization and a native integrations to the CMS. 

We saw the potential of AI early and embraced it to raise quality with our Custom AI Profiles. We backed this with the advanced governance and high security systems that enterprise customers demanded.

Lokalise Vantage

Sahil: Our customers trusted us to deliver across these use cases, so much so that they wanted our platform to be used by all their teams, an intuitive solution that they could deploy without training and without a specialist sitting in the middle of every request. An interface that makes a scaling company faster is the same interface that gets an enterprise team to adopt something new across all their markets. 

To answer this customer need, we've developed Lokalise Vantage, a second workspace on the same platform. 

Lokalise Expert is the workspace customers already know and trust. It is built for product and technical content, short strings, interface text. Lokalise Vantage is new. 

Built for long form content

Sahil: The website copy, campaigns, product pages, and guides your marketing team produces and for the contract and policy content that legal and HR needs in every market. 

These content types are not the same. A homepage headline needs to persuade. A legal notice needs to be exact. A button label needs to fit in the space available. 

Generic localization tools apply one approach to all of them. And that's not efficient and can cost you three times over in your budget, in your quality and in your results. So Vantage routes content by what it is, which language it is going into and how much risk it carries. 

Each type follows its own path. Vantage connects to the content platforms those teams already work in, so localization happens where the work happens. Both Vantage and Export workspaces benefit from the same AI native capabilities, the same brand context, same glossaries and style guides, the same quality controls, governance and audit trails, the same user management and billing. And the value compounds. 

Every translation, every reviewer correction, every piece of performance data feeds back in. 

Each project starts from a better position than the last one. 

Sahil: The more you use it, the better it gets and the more cost effective your localization becomes. So now let me introduce you to Vantage, the new Lokalise workspace for long form content. 

Maailma is a SaaS company. They publish in English and they want to publish in French, German, and Spanish. Until now, that meant sending a file to an agency, waiting two weeks, and getting copy back that needed rebuilding. 

Here is the same job, start to finish, with Vantage. 

It starts in the workspace where your content lives. Name the project, set the base language, and then the target translation languages, choose where the content comes from, easily upload a file or import using our out of the box connectors. Here we are using one HTML file of a blog post. We upload the source file so that the final result arrives with its structure intact, maintaining headings, links, and formatting ready to use. And it's that simple. No engineering ticket, no setup project. 

Before anything is translated, the platform applies what it already knows about this brand. The approved terms and any language level rules. Yours to see, own, and correct. In most companies, this knowledge lives in a few people's heads or scattered across the organization in multiple files. Here, it is a managed asset applied to every job. 

AI alone or AI with human review? 

Sahil: Now the choice that matters, AI alone or AI with human review? With human in the loop, you decide how much gets reviewed, when the reviewer needs to complete the tasks, and you even have the ability to assign multiple reviewers to tasks before a single word is translated. 

Three markets translated. However, it's not just about speed anymore. Anyone can get a fluent translation now. Fluency is the floor. The real question is whether it sounds like your brand and whether it works in the market. 

This is your reviewer screen, not a spreadsheet of disconnected sales. The French heading wraps to two lines. The body runs to four and reaches the bottom of the card, so you fix it here. Ask for a shorter line, choose the one that keeps the meaning and approve. And now you can see the French layout fixed. 

Every change is tracked, what the AI wrote, what a person changed, and when. Side by side, with the source, section by section, so nothing goes out that nobody has actually looked at. For whoever runs this, one view of every task, every language, and what needs attention today. Because the point is not the file, the point is four versions of the same page, each one right for its market. Switch the preview to German and the same page renders again. 

Now filter to what needs a human, fluency, terminology, style, flagged by severity with the reason written out, not a score you have to guess at, an explanation your reviewer can assess and act on. 

Your reviewer gets an email with the task, the scope, and the deadline, and then a person makes the call the brand depends on. 

Here, formal address instead of informal, the kind of judgment that carries a brand in a market and that AI should not make alone.
Reviewed, approved, closed. Every language, every word accounted for. That's Lokalise Vantage.

The reason all of this holds together is the language intelligence layer underneath it. Magnus built it, so I will let him explain it. 

The language intelligence layer

Magnus: I'm Magnus, CTO at Lokalise. I want to introduce our best in class AI capabilities to you. But first, let me start with something that happened to one of our customers. This customer was preparing for a launch in France. With their entire personality as a casual brand, the choice of word is very important. Always the informal version tu, not vous as the model gave them. 

Both are of course perfect French, but one of them sounds like their brand and the other like a tax office. The word choice isn't in the sentence, it's in their brand book. The model in this case did not fail at French. It failed at everything around French. This is my answer to one of the questions I'm often asked. The AI models are excellent now, so why aren't they getting localization right? 

Because the model knows the language. It doesn't know you. Here is what knowing you actually requires. 

Five pillars

Magnus: First, context. Before every job, we retrieve what the model cannot know: your terminology, your tone, your product, how you translated last year. And then we build all of this into the request. For every token of translation we produce, we send a vast amount of tokens of context. The translation is the small part. 

Second, context has to be governed. The biggest challenge is not simply getting this context but in ensuring the right context is provided. It has to be high quality and the right amount. 

Lokalise has invested deeply in ensuring we can provide context that maximizes the quality of the response without oversaturating the model with information. This is managed by our custom AI profiles. Lokalise manages the complexity of the model selection and evaluation. The versions we use are already scrutinized to meet your localization requirements. 

Third, routing. No model wins everywhere. One model is better at German, another is better at marketing copy. And they change places every few months. Connecting to many models is the easy part. Knowing which model wins for your content is the hard part. 

Fourth, judgment. Every translation is scored before a person sees it. People ask a fair question. If a model can find the error, why does the model not avoid making it? For the same reason your code has bugs that your reviewer catches, checking something is different from creating it. Validation of the scoring is critical to this process. It has to be specific to your business and your content. Without this refinement, you won't get the signals that prompt you towards language market fit. 

Fifth, quality decays. When teams do not actively manage context, it starts to reduce the quality of your localization instead of protecting it. A model updates and shifts the tone in two languages, but not in any of the others. Instructions can stop working. Glossaries, translation memory and history all age. 

The feedback loop

Magnus: In most companies, nothing announces any of this, so the problem is never solved. We built a feedback loop that keeps everything contextually relevant. The context is treated as a managed asset, not as a file in a folder. 

Generating a translation is cheap. Trusting it is expensive. That is what we have spent years building.

Most companies are still near the full value of this. Translation and localization are not the same thing, and the gap between them is where the advantage now sits. Not whether your message is accurate in a market, whether it was built for that market and for the channels your customers actually use. This is why we're proud to share the newly expanded Lokalise platform. Now across your whole business, localization stop being a task and becomes a competitive advantage. 

Don't just enter a market. Belong in it.