HOUND EXPRESS

Learn how we helped Hound Express transform its tax operation by automating the classification of thousands of documents with artificial intelligence on AWS, reducing hours to minutes, reducing errors, and achieving 25% savings in operational costs.

THE CHALLENGE
Hound Express faced a critical challenge: automating the tax classification of products for its import and export operations. This process was done manually, with high operating costs and a considerable margin of error. Among the main challenges were:

• Manual processing of 800 to 1000 files per week, with ambiguous descriptions.
• A team of 10 to 12 people dedicated to this task, with a high operating cost.
• Processing times of up to 1 hour per file.
• Error rate of up to 10%, leading to legal risks, customs delays, and lack of scalability.

WHAT WE DID
At Itera we lead the implementation of an intelligent and scalable solution based on generative AI. A completely serverless using key AWS services:

• Amazon Bedrock with Claude 3.5 Sonnet v2: Generative inference engine to accurately analyze descriptions and classify products.
• AWS Lambda + Step Functions: Asynchronous batch flow orchestration.
• Amazon S3: Secure storage of processed files.
• Amazon DynamoDB: Monitoring the health of each processed file.

The solution automatically delivers each product’s tax classification, its status (taxed/exempt) and an auditable explanation per line, ensuring full traceability for regulatory purposes.