Can custom AI really deliver human-level translations?

Language isn’t one-size-fits-all. Every industry, company, and product comes with its own terminology, tone, and context — and that’s where AI customization becomes essential. By tailoring AI to specific domains, localization teams can achieve translations that are not only accurate but also trustworthy and brand-aligned. This level of precision is driving adoption even in compliance-heavy industries like healthcare and legal, where customized AI profiles are now reaching 90% acceptance rates on translations, effectively human-level quality.

In this session, we’ll introduce how custom AI translation profiles work at Lokalise for localization teams. We’ll compare two practical options: fine‑tuned models and a retrieval‑based approach (RAG). Using simple examples with translation memory and past translations, we will illustrate the advantages and disadvantages of each approach, and why we think RAG is the right choice for most teams.

What you will learn:

✅ Why we’re seeing an increase of Custom AI
✅ What are Fine-tuned and RAG models for localization
✅ How do the two approaches differentiate
✅ What are the cost benefit trade-offs

Why this matters:

From generic to specific: Out-of-the-box AI models are trained to be generalists, but localization needs are highly domain-specific. Customization bridges that gap, ensuring translations reflect the company’s tone, terminology, and industry context.
Business impact: Customization means fewer post-edits, lower costs, and faster go-to-market. It also reduces errors that could harm brand perception or user experience.
Scalability: As organizations scale, manual fixes to AI outputs don’t scale with them. Customizable AI ensures consistency across teams, markets, and products.

Who should attend:

Anyone working in global projects and localization efforts, from Product, Content, Marketing, everyone is welcome. No AI experience needed.

Meet your speaker:

Sasho Savkov

Engineering Manager on AIML at Lokalise

Sasho is an Engineering Manager at Lokalise, where he leads the AI translation effort. He brings more than 15 years of experience in AI across academia, business sustainability, and healthtech, and now applies his expertise to the localization industry. Sasho combines deep technical knowledge with a passion for innovation.