← Back to work
Fitness · Europe

European supplier of fitness equipment

From empty product fields to approved content in six languages.

± 2 weeks
to working AI generation
3+ weeks
to writing back to the product information system
82
text fields per product

The question

The supplier wanted to bring new products to its sales channels faster and with consistent content, without handing over control of quality and substance.

Before

The supplier sells high-end fitness equipment to businesses and consumers across Europe. A new product started as a draft in the product information system, with text fields still to be filled. Each product needs 17 Dutch master fields and 65 translated fields in five other languages, including titles, descriptions, bullet points and channel-specific copy. That content had to be ready before the product could go to the sales channels. Every text also had to fit the brand, the product group, the language and the requirements of each channel. The supplier wanted to make this process faster and consistent, while keeping its own editorial control.

What we built

We built a content portal that uses existing product data as the basis for copy and translations. AI checks the product benefits against the specifications, writes the Dutch copy and translates it into five languages. The team manages the writing instructions, brand style, terminology and rules per product group itself, and reviews all content before it is written back to the product information system. For every field you can see which version of the instructions was used, and earlier versions can be restored. Real AI generation was running after about two weeks of building. After just over three weeks, writing back to the product information system was in place too.

How it works

You see what happens at every step.

  1. 01automated

    Collect product data

    The portal finds new products and collects specifications, product benefits, brand and category. A data check shows which source data is missing.

    Role of AI, automation and people: Automation collects the data; the team chooses which products are processed.

  2. 02AI + person

    Check product benefits

    The benefits supplied are checked against the specifications and for duplicates, contradictions and wording. Where benefits are missing, AI can make suggestions.

    Role of AI, automation and people: AI (OpenAI) makes suggestions; the editor accepts or rejects them and decides which points go into the copy.

  3. 03AI

    Write the Dutch content

    The portal generates 17 Dutch master fields from product data, brand style and rules per product group.

    Role of AI, automation and people: AI writes the content. Code handles the fixed HTML structure of the long description and checks such as character limits.

  4. 04person

    Edit and approve

    The editor reviews the copy, adjusts it or has a selected passage rewritten. Changes and earlier versions stay visible.

    Role of AI, automation and people: A person reviews; AI helps with targeted rewrites.

  5. 05AI + person

    Translate per market

    The content is translated into British English, German, French, Italian and Spanish for Spain, with fixed terminology and instructions per market.

    Role of AI, automation and people: AI translates; the team reviews the translations.

  6. 06automated

    Write back to the source

    Once approved, the fields go back to the product information system. Publishing to the sales channels then runs through the existing process.

    Role of AI, automation and people: An authorised employee of the supplier confirms the action. Automation writes the fields back, within a set daily limit.

Result

The supplier has one portal where product content is generated, reviewed, translated and written back: 82 text fields per product in six languages, with its own instructions per brand. The team steers the AI itself through writing instructions, terminology and product group rules, without needing a developer. Every change is traceable, and only authorised employees of the supplier can write approved content back.

1
portal to write, approve, translate and write back
82
text fields per product
6
languages

Yours, and what comes next

On the agreed handover, the supplier becomes the owner of the source code, hosting, database and AI account, including the generated content, prompts and rules. The next step is feeding search data into what the AI writes, so the copy doesn't just read well but gets found too.

Let's get started

Running into a similar process?