How Strix organized 21,000 product descriptions for Melisa pharmacy using AI

Melisa is one of the largest pharmacy and drugstore chains in Poland. As its product range expanded and online sales grew in popularity, the company faced the challenge of organizing a massive database of product descriptions. The data was scattered, inconsistent, and stored in an unreadable format (including old HTML tags, chaotic section layouts, lack of consistency between products, and a large amount of embedded media in the descriptions).

Melisa needed a partner to help structure this information and unlock the full potential of the Ergonode PIM system. Strix took on the challenge, with the practical use of AI playing a key role in the project.

The challenge:

• 21,000 product descriptions created at different times by various people and systems.
  
• Data contained redundant HTML, Word fragments, and images embedded within text.
  
• No consistent structure - for example, it was hard to distinguish between sections like composition and effects in supplements or medicines.
  
• Manual data cleaning would have taken months and required a large team.
  
  

Melisa expected:

• Automatic cleaning of content from codes and formatting.
  
• Division into sections (e.g. description, composition, dosage) depending on the product type.
  
• Preparation of data for implementation in the PIM (Ergonode), which would then power the new online store.

The Strix team developed a proprietary approach that combined AI capabilities with a dedicated tool built specifically for the project’s needs.
instead of performing a simple “clean everything at once” operation, the process was divided into intelligent stages — allowing AI to work more precisely and predictably.

Proper data preparation and defining very specific tasks for AI were crucial. As a result, the model didn’t behave like an unpredictable bot but rather as a well-managed assistant that gradually organized the content into a consistent structure.

To further enhance quality and data safety, control mechanisms were introduced to eliminate the risk of errors or missing information.

The outcome?
thanks to this hybrid approach, the data was cleaned faster, more cost-effectively, and without compromising quality. Melisa gained a product database ready to use in PIM and across multiple sales channels.

• 21,000 product descriptions cleaned and divided into structured sections.
  
• Minimal unit cost - only a few cents per description, plus development time.
  
• Project completed on schedule, enabling Melisa to prepare its product base on time.
  
• Ready-to-use, consistent product catalog that can be easily developed and published across e-commerce, Allegro, and other channels.
  
  

Why it matters for the market

• Practical AI application - Strix demonstrates that artificial intelligence in e-commerce is not theory but a real tool for solving business problems.
  
• Addressing “legacy data” issues - most companies face massive data chaos before implementing a PIM system; AI can be the key to cleaning and structuring it.
  
• Flexibility and time-to-market - with lightweight apps and an AI + development hybrid model, Strix delivers solutions in weeks, not months.
  
• Quality and safety - implemented AI verification through control of completed sections and an additional AI validation model, increasing predictability and minimizing errors.
  
• PIM as the data recipient - while AI played the leading role, integration with Ergonode allowed the client to easily manage and publish new descriptions across multiple channels.
  
  

Conclusions

The Melisa case study shows that:

• AI excels in focused tasks such as cleaning, segmentation, and data extraction.
  
• Strix uses AI smartly - not as a “magic button” but as part of a well-designed process.
  
• Clients gain real business value - organized product data delivered faster and cheaper than with traditional methods.
  
  

Through this project, Strix strengthens its position as a partner that not only implements e-commerce and PIM solutions but also leverages AI effectively in data and process transformation.