From Manual Classification to Intelligent Accuracy: GS1 Mexico’s AI-Powered Transformation with Devoteam

Below, we break down how they moved from manual data entry to intelligent automation with over 95% accuracy.

GS1 Mexico, the organisation responsible for managing global business standards and barcodes in the country, has taken a quantum leap in its digital transformation. Facing a critical bottleneck in the pharmaceutical sector, the organisation partnered with Devoteam to implement an Artificial Intelligence (AI) solution that has redefined operational efficiency.

Below, we break down how they moved from manual data entry to intelligent automation with over 95% accuracy.

The Client: Guardians of the Common Language of Business

With a team of up to 200 employees and serving more than 26,000 members, GS1 Mexico has a critical mission: to ensure data integrity through Syncfonía, its Electronic Catalogue. This cornerstone stores vital information for millions of products, guaranteeing regulatory compliance in highly sensitive sectors such as logistics and retail.

The Challenge: The Bottleneck in the Pharmaceutical Sector

GS1 Mexico’s Data Verification Service (SECODAT) was facing a major obstacle. The classification of pharmaceutical products and the assignment of codes was a fully manual process, leading to three critical issues:

1. Operational Slowdowns: Significant delays due to human dependency.
2.Error Prone: Manual transcription is naturally susceptible to mistakes.
3.Ambiguous Data: Difficulty interpreting non-standardised descriptions (e.g., “CLORANFENICOL UNG OFT 5G”).

The existing system was not scalable and hindered the organisation’s agility in responding to the growing volume of medical data.

The Objective: Automating Integrity

GS1 Mexico was not looking for a quick fix, but a foundation for the future. Their goals were clear:

• Full Automation: Eliminating manual dependency.
• Extreme Accuracy: Increasing precision in product code assignment.
• Empowering Talent: Enabling experts to move from “data entry” to “quality validation”.

The Solution: “Farma Data Extractor” API

Devoteam designed a serverless architecture hosted on Google Cloud Run, inspired by the Google Agent Development Kit. The solution splits the cognitive load into two specialised AI agents (Multi-Node Architecture):

1. Node 1: “The Eyes” (High-Fidelity OCR)

Using Google GenAI (Gemini), this agent processes images to transcribe all visible text, preserving the original layout with near-perfect accuracy.

2. Node 2: “The Brain” (Extraction and Parsing)

This agent combines the transcribed text with visual context. It uses Pydantic schemas to enforce a strict, structured JSON format, removing ambiguity entirely.

Why this architecture? By separating the act of “seeing” from that of “thinking”, the solution achieves far greater accuracy than monolithic models. In addition, being serverless means it scales automatically according to demand, with no cost for idle infrastructure.

Technology Stack:

• Google GenAI (Gemini): AI engine.
• Google Cloud Run: Scalable hosting.
• Python & FastAPI: High-performance logic.
• Pydantic: Strict data validation.

Results: Measurable Speed and Accuracy

The implementation of the “Farma Data Extractor” immediately transformed GS1 Mexico’s operations:

• >95% Accuracy Rate: Iterative validation that exceeds previous human standards.
• From Hours to Seconds: Complete removal of manual data-entry bottlenecks.
• Automatic Scalability: Handles workload spikes without additional infrastructure.

GS1 staff have shifted from being “data entrants” to data auditors, significantly improving the quality of the final service.

Next Steps: Expanding into the Nutritional Sector

The success achieved in the pharmaceutical sector has paved the way for expansion. GS1 Mexico is already adapting this AI architecture for the nutritional sector, with the goal of extracting complex tables (fats, sugars, sodium) from food labels, further strengthening its leadership in digital transformation.