Straight answers, before you have to ask

It usually is faster — but even companies building AI translation tools still put a human editor on the output. Industry research puts that figure at 90–98% of companies using AI/MT still relying on human post-editing (RWS Translation Technology Insights, 2025). Raw AI output is consistent at scale but still misses tone, context, and register — the exact things a human review catches before something ships to a customer.

Pure human translation is thorough but slow — often too slow for a real product release calendar or a marketing campaign with a fixed launch date. Voxiis uses AI for the first-pass speed and a human reviewer as the quality gate on every single segment — you get the turnaround of AI with the reliability of human review.

We grade every finding by severity (Critical, High, Medium, Low) using an error-typology approach based on industry-standard frameworks like MQM (Multidimensional Quality Metrics) and LQA (Linguistic Quality Assessment). You get an English-readable report showing exactly what was checked and what was changed — not just a translated file and an invoice.

Yes — see our Sample QA Report page for a full illustrative example, including the severity legend we use.

No. Client content is used to produce your translation, not to train models. If you need this in writing for procurement or legal review, we provide it as part of onboarding.

We align with SOC 2 and ISO/IEC 27001 practices and operate under signed Data Processing Agreements (DPAs) for any project involving personal data, consistent with GDPR Article 28 requirements for data processors. Ask us for current certification status and our sub-processor list during scoping.

No. We work inside the tools you already use — Lokalise, Contentful, XLIFF files, or your CMS directly. Nothing to migrate, no new platform for your team to learn.

Your translation memory, glossary, and content stay yours — delivered in standard, open formats. There's no proprietary platform holding your data hostage, and no penalty for moving on if you ever need to.

It depends on volume and language count, but our AI-speed first pass means we can often absorb a launch-scale volume that would take a human-only team six to eight weeks — condensed into days, with human QA still applied to every segment.

We don't publish a flat per-word rate because pricing depends on your volume, language count, and how much ongoing work you need — see our Pricing page for how each model works.

Send us a page or file via our Contact page for a free quality audit — no commitment, just a look at where your content actually stands today.