International growth · Brazil · Localization · Consumer apps

Launching a US app in Brazil: what AI translation still can't do for you

Every few weeks someone asks me some version of this: "We have a US app doing well. AI translation is basically perfect now. Can't we just turn on Portuguese and launch in Brazil?"

I understand the logic. The tools really are impressive. I'm Brazilian, I work in Portuguese and English every day, and I use AI translation all the time. It's fast, it's mostly correct, and it saves hours.

But I led Hallow's growth in Brazil from around 300 thousand downloads to more than 5 million, number one on the App Store and Google Play in the country two years in a row. And I can tell you, translation was maybe the easiest part of that story.

I wrote before about why international growth fails before translation even begins. This time I want to talk specifically about AI, because a lot of companies now think the tools solved localization. They solved a part of it. Here's what's still on you.

Correct and right are two different things. AI will give you grammatically correct Portuguese. But Brazilians will feel immediately if the text sounds like it came from outside. The tone is different. We're warmer, more informal, we use diminutives, we talk to the reader like a friend. A sentence can be perfect and still sound cold, or sound like Portugal, or sound like a manual. That gap is small on a word level and huge on a trust level.

Culture isn't in the dictionary. In a faith product, for example, you can't just translate the devotions that work in the US. Brazilian Catholicism has its own heart, Nossa Senhora Aparecida, popular devotions, the way people pray in their families, the music. If the content doesn't feel like it was made for us, people use it once and go back to what they know. AI can translate a prayer. It can't tell you which prayer your Brazilian user's grandmother taught her.

Trust is built through local institutions and people. One of the moments that marked Hallow in Brazil was our partnership with the Christ the Redeemer Sanctuary in Rio, projecting a message of prayer for Pope Francis on the statue while he was in the hospital. It went viral, more than 100 million views. No translation tool gets you that. That came from relationships, from understanding what that monument means to Brazilians, from showing up with respect.

The business model needs localizing too. Pricing in reais, how people pay, how they feel about subscriptions, how sensitive they are to price, which channels they actually use to talk to brands. In Brazil, WhatsApp is not just a messaging app, it's where life happens. A US playbook copied and translated will miss a lot of this, and AI won't flag it because you didn't ask.

Your team needs real local people. You need someone who can read the AI translation and say "no, nobody talks like this." Someone who knows which influencer is loved and which one will get you canceled. Someone who can sit with a partner and build a relationship in Portuguese, with all the coffee and conversation that implies. That person makes the AI useful instead of risky.

So where does AI actually help when you're expanding to Brazil? A lot of places, honestly:

  • First drafts of content at scale, which a native speaker then adapts
  • Transcribing and translating interviews with local users so the whole company can hear them
  • Testing many variations of messages quickly to see what resonates
  • Summarizing local reviews, comments and support tickets to find patterns
  • Helping a small team look much bigger than it is

Used like this, AI is a gift. It lets a small local team do the work of a much bigger one, and gives headquarters a window into a market it doesn't speak.

The mistake is thinking that because the words are right, the market is ready. Brazil is a huge, generous, loyal market. When Brazilians love a brand, they bring their whole family. But they can feel when something wasn't made for them. And that feeling doesn't show up in any translation score.

If you're thinking about Brazil, start with people before tools. Then let the tools help those people go further. That's how I'd do it again tomorrow.

A note when there’s something worth sharing.

New articles, things I’m building and the occasional lesson from the work. I’ll write when I have something worth sending.