Retresco

Retresco has been doing language technology from Berlin since 2008, which in this field is close to prehistoric and means it was automating text generation long before the current wave made it fashionable. That longevity shows in where it is deployed: publishing, finance and e-commerce, where text has to be produced on a schedule and be correct rather than merely fluent.

Its natural language generation turns structured data into editorial copy — match reports, financial summaries, weather, property listings, product text — the categories where the underlying facts are already in a database and a human writing them adds transcription rather than judgement. Around that sits language analysis for tagging, structuring and finding content, which is what a publisher with a large archive needs almost as much as generation.

Retresco GmbH employs more than seventy people across software development, linguistics and editorial, which is an unusual mix and the reason its output reads like it was edited rather than assembled.

It operates under German jurisdiction and GDPR, hosted on ISO 27001-certified servers in Germany, with enterprise pricing on request. The limits are the ones that come with that positioning: sales-led with no self-serve entry, aimed at organisations with volume, and considerably more than a marketing team wanting help with a newsletter needs.

What Retresco does well

• Automating editorial text since 2008
• Generation plus language analysis for large archives
• Linguists and editors on staff, not only engineers
• German company under GDPR
• Proven in publishing, finance and e-commerce

Where Retresco falls short

• Enterprise pricing on request, no self-serve entry
• Aimed at volume rather than occasional writing
• Setup requires structured source data
• Overkill for a marketing team

Standout feature. Eighteen years of automating text before it was fashionable — and linguists on staff, which is why the output reads edited.