The Workflow Process:

1. Data Input: Facts and content are provided to the workflow
2. Text Preparation: Content is split into discrete sentences for granular analysis
3. AI Verification: Each sentence is compared against facts using "bespoke-minicheck" model
4. Discrepancy Detection: The model returns simple Yes/No verdicts on factual alignment
5. Result Aggregation: Non-factual statements are collected and organized
6. Summary Generation: Qwen2.5 model produces an executive summary with key metrics

This modular architecture enables precise identification of problematic content while maintaining processing efficiency—analyzing hundreds of statements in minutes rather than hours.