"Using Google Translate can lead to some serious errors, especially when words have multiple meanings, which is often the case in fields such as law or engineering," says Samantha Langley, a former lawyer who is now a court-approved French-to-English legal translator based in Meribel, France.
"It has taken decades of research to create a framework of algorithms designed to recognise patterns in the same way as the human brain - a neural network," says Andrew Ochoa, chief executive of US start-up Waverly Labs, which produces translation earpieces.
"Combining that with speech recognition technology has allowed us to make a huge leap forward in terms of accuracy."
There's no doubt that CAT tools have taken some of the hard grind out of text translations like instruction manuals or questionnaires, says Milan-based Paola Grassi, a professional translator for Wordbank, a global marketing and translation agency.
"Survey contents are among the most repetitive ones and a good CAT tool can hugely speed up the process," she says.
For meetings and conferences, wearable translators like Waverly's are undoubtedly popular. But even this new generation tech, which combines speech recognition neural networks and internet-based translation engines, has limitations.