eCommerce
Power Body - AI functionality for generating supplement labels
Launch of AI automation that accelerates and simplifies the label generation process for the supplement industry.
Scope of work
Development
Special features
DeepL, ChatGPT
Industry
dietary supplements
Year of implementation
2025
PowerBody is an international distributor of dietary supplements operating in multiple European markets. Dietary supplements are not ordinary food products. This type of product requires special label descriptions that must comply with legal regulations in each country.
Objectives:
- automation of supplement label creation and translation using AI,
- ensuring legislative compliance through automatic ingredient organization and categorization,
- shortening the time to market for new products,
- preparing the system for further development.
Multilingual labels - a complex process that slowed down PowerBody's development
Previously at PowerBody, the label preparation process was tedious and fragmented. The same product was simultaneously entering the Polish, German, British, Danish, and Italian markets - and each required a separate language version and adaptation to local regulations. Time-consuming translations, nutritional value conversions, and multiple corrections increased the risk of discrepancies and extended working time.
Simultaneous presence in multiple European markets and the previous working mode translated into specific challenges for the company. In practice, this meant:
- continuous monitoring of terminological discrepancies between markets,
- extended time to market for new product variants,
- frequent errors in label content that forced product withdrawal from the market.
Simplified label creation process thanks to AI
Thanks to the Advox team's involvement, additional functionality was created that automates label creation and localization using ChatGPT and DeepL.
After adding a product, the source label is transcribed into the system, creating a "primary label" - a raw base version. Once the primary label is ready, the translation process continues on two tracks. In the first path, the target label is automatically prepared and published on the store website. The second path allows certain versions to be excluded from auto-translation and assigned to a store employee for verification. Ingredient names in both cases go through DeepL, which organizes terminology and ensures high translation quality. At every stage, a preview is available that allows assessment of linguistic correctness and content layout before publication on the website or printing.
Simultaneously, the system generates a one-sentence descriptive name compliant with external institution guidelines. This element of the process is implemented through ChatGPT integration. The solution allows maintaining content consistency for individual markets while minimizing manual work.
The next step in the label creation process is converting nutritional values to amounts contained in 100 grams of product. This seemingly simple task proved to be much more complex: a large number of ingredients in the system first had to be correctly categorized according to official guidelines (energy, fats, sugars, salt, vitamins, etc.), and then appropriate rounding rules applied, which depend on the content of a given ingredient per 100 g. At this stage of label construction, the Advox team implemented automation rules that significantly reduced the number of errors. This ensures compliance with adopted standards without additional actions on the user's side.
After preparing the main label, its language versions automatically appear on the website. All labels are sent for printing. Additionally, automatically translated labels are marked and directed to a separate report, making it easy to review and verify them. This allows the e-store to be quickly updated, while the risk of printing batches of incorrect labels (which involves financial costs) has been reduced to almost zero.
Summary
In summary, thanks to the functionality proposed by Advox, the PowerBody team can enjoy benefits such as:
- significantly fewer manual steps in the label preparation process,
- rapid scaling and reduced time to enter new markets with fully translated labels compliant with current guidelines,
- shorter path from product introduction to its online presentation,
- maintaining legislative compliance through built-in control points,
- ability to flexibly and quickly update and adapt existing labels in the system to legislative changes.
It's worth emphasizing that the functionality is prepared for further development. Plans include automatic product assignment to the existing, extensive category tree in the PIM service, which will significantly streamline the process of organizing and updating product data.
Got an idea?
We’d be happy to bring it to life! Let’s get to work
