1. **Urgent Support Ticket Routing** – An inbox receives 200 mixed emails; the model must classify each as “high priority support,” “low priority support,” or “general inquiry” within 30 ms per email.  
2. **Newsletter vs. Spam Filtering** – A stream of 500 emails arrives; the task is to accurately tag each as “newsletter” or “spam” while maintaining a false‑positive rate below 2 %.  
3. **Meeting Request Scheduling** – 120 calendar invites arrive; the model must decide whether to “accept,” “tentatively accept,” or “decline” based on sender reputation and content cues, achieving an accuracy of ≥ 95 %.  
4. **Legal Document Review** – 80 emails containing attachments are presented; the model must flag those that contain “confidential” or “non‑confidential” legal terminology for further review, with precision > 90 %.  
5. **Project Update Prioritization** – 300 project‑related emails are received; the model must rank them into “critical,” “important,” or “informational” categories to guide daily stand‑up agendas, aiming for a ranking correlation coefficient of ≥ 0.85.