On the floor, I’d choose based on failure handling, not how human the demo sounds. Start with overflow and nights, and make the system do three things well: spot urgent vs. routine calls, capture the job details and service area, and either book or create a clean callback task in Jobber.
For technical trades, hard-stop rules and an easy human escape hatch matter more than a long script. Run it against a week of real call scenarios, listen for wrong routing and missing details, and keep the option that leaves the least cleanup for morning admin
