Manual audit and review

For Premium Classify users, Zonos will manually audit, review, and correct a subset of your Zonos Classify requests to constantly and consistently improve the accuracy of your classification results.

Audit and review process 

Classify is built on a multimodal AI system powered by billions of training records and advanced algorithms, enabling it to deliver accurate and reliable results from the start for the majority of customers. However, we recognize that certain scenarios—such as low-risk business models or low-quality input data—pose unique challenges. To address these challenges and ensure even higher accuracy, we have developed a comprehensive manual review process:

• Automated flagging: Classifications are automatically flagged if they fall within the lowest X% of confidence scores, depending on your use case and agreement.

• Expert manual review: These flagged classifications will then be manually reviewed by our expert classifiers.

• Continuous learning and model improvement: The manual classifications are then integrated into our AI model, enabling it to learn and improve over time.

• Manual classification retrieval: These manually classified items are then stored in a specialized table. For new classification requests, the Classify flow will first check this table for an existing manual classification based on the item's unique identifier (item key). If a match is found, it is used. If not, the request progresses through the standard Classify flow.

By integrating advanced AI capabilities with manual expertise, our approach ensures top accuracy for even the most challenging classifications. This combination allows you to benefit from the speed of automated processes and the precision of expert human review, leading to consistently reliable results.

Note: The manual audit and review offering is automatically implemented for Classify Premium users. Learn more about our Classify tiers.