The problem
Some work, such as internal reports and gist translations, is agreed with the client as machine translation plus human post-editing, but many small agencies still translate it from scratch. At 8,000 words a month and 400 words an hour, that's 20 hours of in-house time on text that a good draft would speed up.
What the AI does
DeepL Pro translates text and documents and lets you set glossaries so a client's terms are always translated the same way. DeepL's page lists no data training on its plans, and its Team plan includes 5 glossaries and 20 file translations a month per user. For jobs the client has agreed can be machine translated, your linguist uploads the document with the client's glossary, then post-edits the draft fully before review and delivery.
A worked example
A translation agency in Bristol with 2 in-house linguists handles about 8,000 words a month of agreed post-editing work: internal reports and gist translations for 4 corporate clients. Both linguists use DeepL Team, with a glossary for each of those clients.
Before
8,000 words a month translated from scratch at 400 words an hour: 20 hours
After
8,000 words post-edited from DeepL drafts at about 700 words an hour: about 11.4 hours
What it's worth
About $343£258€307A$495C$485 a month in time saved: 8.6 hours (20 down to 11.4) at an assumed $40£30€36A$58C$56 an hour. DeepL Team is €24.99 (about $28£21€25A$40C$39) a user a month billed annually, so 2 users is about $57£43€51A$83C$81 with VAT included. Prices were shown in euros on DeepL's page; check the price in pounds at checkout.
How to set it up
- Agree in writing with each client which jobs may use machine translation with post-editing, and at what rate.
- Start DeepL's 30-day free trial on the Team plan for the linguists who post-edit.
- Build a glossary for each of those clients from their approved terms.
- Post-edit every draft fully, then have a second linguist review before delivery, as on any other job.
- Keep certified, legal, medical and marketing work out of this process unless the client asks for it and you're confident in the result.
What goes wrong
- Machine translation without consent can breach your client contract. Get agreement before using it on a client's text.
- Team includes 5 glossaries and 20 file translations a month per user. More clients or files needs Business at €49.99 (about $57£43€51A$83C$81) a user.
- Fluent output can hide wrong meaning, a missing negative or a mistranslated number. Post-editing means reading every sentence against the source.
- Keep personal data out unless your contract with DeepL and your client's terms allow it.