explore
swipe
Please turn off low power mode for a smoother experience.
Artificial Intelligence

Text at volume, without becoming arbitrary

We produce and extend text with AI: product descriptions, variants for different channels and translations, following fixed rules for tone of voice and terminology.

Let's connect 🚀

What this is about

AI writes in minutes what used to take days. That is precisely the risk: without rules you get text that sounds correct and belongs to nobody. The effort shifts from writing to checking, and without rules the checking becomes more expensive than the writing.

Rules before tools

Tone of voice, permitted terms and sentence lengths have to be written down before anything is produced. Otherwise every text sounds different.

Strong at volume and repetition

The benefit comes with product texts, variants and translations. For a single lead article the effort rarely pays off.

Facts come from your data

We pull dimensions, prices and properties from the PIM or the shop instead of letting the model guess them.

Writing quickly is easy. Sounding like you is the work.

Situations that typically bring clients to us

  • Product descriptions are missing for hundreds of items.

  • Texts sound like everyone else, just not like your own brand.

  • Every channel needs a different length of the same text.

  • Content should appear in several languages, without a translation agency.

  • The editorial team should publish more often without growing.

  • Existing texts are outdated and nobody gets round to revising them.

  • None of this sounds like you?

    Then tell us what you are actually dealing with.

    Get in touch

What we actually do

Writing down the rules

Tone of voice, form of address, permitted and forbidden terms, documented and checked against examples.

Product and category texts

Descriptions in large numbers, built from your existing product data rather than from guesswork.

Variants per channel

One text turns into a short version, an ad text, a newsletter version and a description for search.

Translation and terminology

Content in further languages with consistent technical terms instead of labels that keep changing.

Reworking existing text

Updating and shortening existing texts or shifting them to a new tone of voice, without rewriting everything.

Release with labelling

A process in which suggestions are recognisable as such and are checked before publication.

A matching journey

Screenshot Deutsches Handwerksblatt

Deutsches Handwerksblatt

AI-assisted suggestions for titles, descriptions and subheadings directly in the CMS, following the publisher's guidelines.

View journey

How we work

  • 01

    Clarify what is needed

    Which texts are produced how often, and where does it cost the most time today?

  • 02

    Set the rules

    Tone of voice and terminology are set down in writing and checked against real examples.

  • 03

    Test on a small scale

    We start with a manageable volume and review the results together with your editorial team.

  • 04

    Bring it into everyday work

    Whatever holds up is integrated into the existing system, so nobody has to switch between tools.

AI does not replace an editorial team, it takes the repetition off their hands. Tell us which texts cost you the most time.

Let's connect 🚀

More journeys

Two projects where AI helps with the writing without replacing the editorial team.

Deutsches Handwerksblatt project

Deutsches Handwerksblatt

AI-assisted suggestions for titles, descriptions and subheadings directly in the CMS, following the publisher's guidelines.

View journey
PUKY project

PUKY

A shop with automated product imports from the PIM. Clean product data is the basis of any reliable content production.

View journey

Related services

Text is not written for its own sake. Images belong with it, the integration into your own system and the question of whether the content is actually found.

Frequently asked questions

Writing quickly is easy, sounding like you is the work. AI speeds things up, while the editorial team and the facts stay guided.

Does the AI at dxm write the texts on its own?

No. Before production we set down writing rules – tone of voice, form of address, permitted and forbidden terms, sentence lengths – and check them against examples. Only then is text produced at volume. The effort shifts from writing to checking, and without rules the checking becomes more expensive than the writing. Suggestions are recognisable as such in the process and are approved before publication.

What is this particularly worth it for?

For volume and repetition: product and category texts for hundreds of items, variants of the same text per channel – short version, ad text, newsletter, description for search – translations with consistent terminology and the reworking of outdated material. For a single lead article, or a text meant to convey a position, the effort rarely pays off.

How do you prevent invented details?

Facts come from your data, not from the model: we pull dimensions, prices and properties from the PIM or the shop instead of letting them be guessed. Statements that cannot be backed up do not go into the text. On top of that, a person approves before anything goes live. With large volumes we additionally spot-check against the source data.

Does this work in several languages too?

Yes, with two caveats. We set technical terms in a terminology list beforehand, so a different label does not appear in every language. And for languages where nobody internally can judge with confidence, we recommend a check by native speakers or a partner. We do not use it as blind translation without any control.

Too many texts, too little time?

Tell us which texts cost you the most time. We will look at where AI genuinely helps.

Let's connect 🚀