Document Processing · Risk Management
The Customs Committee of Uzbekistan needed to process growing volumes of customs declarations that contained complex free-text goods descriptions, but manual review slowed clearance, introduced classification risk, and limited the agency’s ability to identify compliance issues efficiently.
Reported impact
The AI-API cut declaration processing from hours to seconds, automated tens of thousands of declarations, improved compliance and revenue collection, and gave Uzbekistan a scalable public-sector example of AI-driven customs modernization.
IBM Expert Labs and IBM Client Engineering worked with the Customs Committee of Uzbekistan to turn customs declaration handling into an API-driven AI workflow built for scale. Using IBM Enterprise Design Thinking, the teams defined priority use cases and then built an MVP on IBM watsonx.ai with IBM watsonx.data as the data layer for efficient access across distributed sources. The solution uses a Llama 3.3 large language model to extract structured information from declaration text, classify products automatically, and feed those outputs back into the country’s already digital customs environment through an interface that employees can adopt quickly. By pairing generative AI with scalable data access, the Committee created a foundation for predictive classification, transport-cost calculation, and future customs-control use cases without rebuilding the wider clearance infrastructure. IBM’s hands-on delivery also kept the project aligned to operational and compliance objectives, not just technical experimentation.
Declaration text enters the AI-powered API, Llama 3.3 on watsonx.ai extracts structured fields and classifies goods, watsonx.data supports access to the underlying data sources, and the results flow back into the existing digital customs system so declarations can be cleared, re-registered, and reviewed for compliance in seconds rather than hours.
“In 2024 alone, more than 80,000 declarations worth USD 5 billion were completed without the participation of employees.”
“The solution is built on a powerful API that uses Llama 3.3 large language model (LLM) to extract structured data from text fields and accurately classify products.”
“Using the IBM watsonx.ai AI studio, the teams jointly developed an MVP with advanced generative AI and machine learning technologies.”
“reducing the previous 2–3 hours of processing time to just a few seconds.”
“the solution integrates the IBM watsonx.data data store”
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
AWS Sample Apr 15, 2024 InfoCert Training and deploying these models was a lengthy, costly process, which limited access to this set of capabilities to larger companies. While InfoCert was searching for a way to streamline such ML integrations, AWS… Additionally, InfoCert expects to reduce financial overhead by 80 percent for startup, development, and… Amazon Bedrock 
Databricks Sample Jun 13, 2023 International Finance Corporation (IFC) As emerging markets seek to attract more capital to spur economic development, investors are considering ESG issues such as impacts on local communities, value of goods generated, and greenhouse gas emissions when… 1,200 ESG risk terms Agent Bricks · Delta Lake · Lakeflow Jobs 
IBM Sample Apr 6, 2026 United Nations Development Programme UNDP needed more granular, forward-looking analytics to help governments and development partners decide where clean-energy investments would produce the greatest equity, development, and climate-resilience impact. 1 km resolution IBM Cloud · IBM Environmental Intelligence · IBM watsonx.ai 