You may know this as adjudication, human in the loop, human + AI collaboration, exception handling, or review of what your quality estimation flagged.
What it is
Your engine translates. A model scores every segment for how likely it is to be wrong. An automated editing pass resolves what it can. Only what is left reaches a person.
That residue is what we work on — and by definition it is the hard part. Ambiguity the model could not resolve. Terminology that has to match a document nobody gave the machine. Formality and register the source language never stated. Content where being confidently wrong is expensive.
What comes back
Every disposition carries a named reviewer and a timestamp
That record is part of the deliverable, not an internal artefact — in regulated content it is often the part your own client is buying.
Where translation comes in
Where the machine's version cannot be edited into shape, we translate the segment properly. It happens more often in some languages than others: in the 2025 evaluation, professional human translation beat every machine system tested into Japanese, Icelandic, Estonian and Egyptian Arabic. Those are the pairs where retranslated is regularly the correct disposition, and where a per-word post-editing discount is hardest to justify. More on that under human translation.
How we work with your automated checks
We assume they exist. Buyers increasingly run quality estimation, an automated review pass, or both, and route only what falls below a threshold to a person. That is a sensible design and we are built to sit at the end of it. Send us the exceptions.
Where it matters, we will also tell you when we think a threshold is set wrong — which is a conversation the party that configured the threshold cannot easily have with you.
Independence
We do not sell a translation engine. We do not sell a platform. On the work we review, we were not the producer. An assessment written by the party that produced the output is weaker evidence, and regulated buyers have started saying so out loud.
Where we have produced the content ourselves, we say so and we do not also certify it.
What it costs
Per segment, or per hour. Not per word — per-word pricing assumes machine output quality is constant, and it is not. A clean segment and a badly mangled one would pay us the same, which is a pricing model that penalises exactly the work you are sending us.
Delivered, or staffed
Send us the work and we return the deliverable described above. Or place our reviewers into your own process and tooling and run them under your rubric, your schema and your brand. Same people, same documentation, different contract. Tell us which you want and we will price it that way.
Need the evidence as part of the deliverable? See documentation and evidence.