Why human quality review is still needed
The knitting pattern example says a lot about why human quality review is still needed. When AI was asked to translate the Swedish word “varv” into German, it lacked the context needed to decide whether it referred to knitting in the round or knitting back and forth. The result was a translation that was technically correct — but still wrong in this particular context.
These are exactly the kinds of nuances that human language reviewers catch. AI can work quickly and efficiently, and both prompts and glossaries can be improved over time. But when large volumes of content are translated, small slips can quickly have a big impact on both customer experience and brand. That’s why human quality review is still an important part of the process.
Light proofreading
A quick human check of grammar, consistency and obvious errors. Ideal for high-volume content where AI quality is generally high, but where a final human pass adds extra peace of mind.
Full post-editing
A thorough human review and refinement of AI-generated content — with adjustments to language, tone and terminology. This is our most popular QA level for product descriptions and
marketing content.
Localisation review
Beyond linguistic accuracy, we make sure the content also works culturally in the local market. Does the content resonate with the target audience? Are idioms, references and tone of voice properly adapted to the market?
Our linguists always work into their native language and have a solid understanding of the markets they work with.
Improving AI prompts through feedback loops
Our QA doesn’t just correct errors — it also helps improve the AI workflow over time. Our linguists identify recurring patterns and help refine glossaries, tone of voice and Lexi prompts.
That way, the system gradually gets better with every review.
How it works
1. Content is translated with AI via your
configured Lexi solution.
2. The content is flagged and categorised for
human review.
3. A specialist reviewer checks, comments and corrects.
4. The feedback is structured and used to improve the AI prompts.
5. Quality improves step by step, without you having to do a thing.
Choose the quality level that matches your needs
Essentials
AI translation with automatic quality scoring and simple spot checks. Best suited for high-volume, low-risk content.
Professional
Full post-editing by a specialised language reviewer.
Recommended for customer-facing product and marketing content.
Premium
Complete human localisation review, contextual website annotation and feedback for prompt improvement. For brand-critical content where quality is non-negotiable.
Frequently asked questions
How much slower is human QA compared with AI only?
Full post-editing normally takes 24–48 hours for standard volumes. We’ll always give you a realistic time estimate before the project starts.
Can I request human review only for specific languages?
We’ll help you find the right setup.
What expertise do your reviewers have?
specialist knowledge in the relevant content areas.
We don’t use generalists for specialist content. For e-commerce, that means reviewers who understand product names, SEO and conversion-focused copy.
Does the feedback loop really improve AI quality over time?
Can I choose human review only for certain types of content?
With our integrations, this can be managed automatically.