Quick answer: for everyday business text — customer emails, product and service descriptions, internal notes, FAQs — AI translation is a genuinely useful first drafter, and usually faster than starting from scratch. For documents with consequences — quotes, contracts, terms, anything an authority or court might read — treat the output as a draft for a competent person to check, never as a finished translation. The catch with AI translation is that fluency hides errors: the text reads well whether or not it says what you meant. Below: where it helps, where the risk concentrates, and a five-point check to run before anything reaches a customer.
The problem is not bad translation. It is confident translation.
Old machine translation was easy to distrust — you could see the seams. Modern AI translation reads like polished writing by a fluent speaker, and that is exactly what makes review harder. A sentence can be grammatically beautiful and still carry the wrong meaning, the wrong register or the wrong regional flavour. The reading experience tells you nothing about accuracy, so "it reads well" cannot be your quality check. Something has to replace it — that is what the five-point check below is for.
Where AI translation genuinely helps
- Customer emails and replies. When a customer or supplier writes to you in the other language, an AI draft gets you to a polite, clear reply quickly. You check names, dates and any commitment before sending.
- Product and service descriptions. First drafts of website copy, brochures and listings. For a UK business selling to Portuguese customers — or the reverse — this is where the speed difference is felt most.
- Internal documentation. Process notes, checklists and training material for a mixed-language team, where a small imperfection is cheap to fix.
- Understanding incoming text. A contract or official letter in the other language, translated so you can follow it before a professional looks at it. Useful for orientation — not a substitute for that professional step.
- Terminology exploration. Asking how a term from your trade is usually rendered in the other language, then confirming with industry sources or a professional.
Where the risk concentrates
- Documents with legal or commercial effect. Quotes, contracts, terms and conditions, complaints, anything formal to a regulator or authority. Errors here become commitments or disputes. If the recipient is an authority, court or notary, ask them first what they accept — many require a certified or sworn translation, which AI output is not.
- Numbers, dates and currency. Date formats flip between conventions, and a slip in a digit survives every grammar check because it is not a grammar error.
- Regional variants. European Portuguese and Brazilian Portuguese are one written language with different vocabulary, spelling decisions and register (equipa/equipe, registo/registro, ecrã/tela). British and American English differ in similar ways. An AI assistant does not always guess your target variant from context — say which one you want.
- Your own terminology. Product names, service names and the technical terms of your trade. Letting a tool improvise here is how one document ends up calling the same service three different names.
- Tone and register. Portuguese business writing tends to be more formal than its British equivalent; a cheerful direct translation can sound oddly casual to a Portuguese reader, and an over-formal Portuguese draft can feel stiff in English.
Five checks before you publish
- Names, numbers and dates, against the source, character by character. They carry the commercial meaning and are the easiest place for any translator — human or AI — to drift.
- The regional variant is the one you intended. PT-PT for Portugal, PT-BR for Brazil; en-GB or en-US as appropriate. If you serve both markets, that is two documents, not one compromise.
- Your terminology is consistent. Make a short list of your business terms — product names, service names, trade terms — and decide once how each appears in the other language. Keep the list next to the tool.
- The register matches the audience. Read it as the recipient would: is the level of formality what a business reader in that country expects?
- The consequences test. Ask: if the recipient believed the least accurate sentence in this document, what would it cost? If the answer is money, a commitment or a dispute, the document needs a professional translator, or at least a reviewer fluent in both languages, before it leaves.
One small example of how it goes wrong
(Illustrative example, not a client story.) An English draft says "We should be able to deliver by Friday" — a cautious estimate, and the sentence you want the client to read. The Portuguese draft comes back as "Conseguiremos entregar até sexta-feira" — "we will be able to deliver by Friday". The hedge vanished in translation. No spelling mistake, no broken grammar, nothing to alert the reader — just a quiet hardening of a commitment, in exactly the sentence the client will quote back to you. Modality — should, may, intend to — is where translation drift changes meaning without touching a single noun, and it is why check 5 asks what the least accurate sentence would cost you.
When AI translation is not the right tool
- The document has legal effect, or the receiving authority requires certification.
- The text is brand-critical: a launch campaign, a homepage, a tagline. Nuance is the product here; use a fluent native speaker.
- Safety or compliance instructions where a mistranslation could cause harm.
- You cannot check the output and nobody else in the business can either. If nobody can verify it, that is precisely the moment to pay for the review.
The workflow that keeps the speed and removes most of the risk
AI drafts. A person who knows both languages checks the draft against the source, focusing on the five points above. Terminology decisions go into a short glossary that gets reused, so consistency improves with every document. That loop — draft, check against the source, keep the decisions — is the same discipline that makes any AI-assisted work trustworthy, and on a routine email it takes minutes.
Try the check on one real document
If you work across English and Portuguese and want to build this checking habit on your own material, that is what the Personal AI Session is for: 60 minutes online, one-to-one, on one task you prepare and anonymise in advance — a customer email, a service description or a document you translate regularly. We work through it together, including where AI translation helps you and where it should not be trusted. It costs £75 per session (89 € in Portugal), the final total for the scope listed, and it includes a short written recap within two working days plus a follow-up clarification email thread. Requesting a session creates no booking or payment obligation.
Still deciding? Book a free 30-minute call — no preparation needed, and if training is not the right fit, we will say so.
Related reading: Should AI draft your quotes and proposals? — the same discipline for documents that commit you to prices — and does your small business need an AI usage policy?.
Last updated: 4 November 2026