Three developments in the sources make this a useful moment to move beyond demonstrations:
- Operational integration is becoming a concrete procurement question. ServiceTitan documents booking, rescheduling, configurable escalation and job-level identification of voice-agent bookings. These are functions buyers can test against their own dispatch rules, rather than infer from voice quality. (help.servicetitan.com)
- Evaluation can inspect more than transcripts. LiveKit’s documentation distinguishes text simulations from audio simulations and allows additional grading against the agent’s final system state. That supports a more demanding question: did the conversation produce the correct record? (docs.livekit.io)
- Management responsibility remains central. Brad Cleveland’s ICMI article, updated on 15 October 2026, frames automation scope, accountability and changing human work as leadership decisions, not merely technology implementation. (icmi.com)
These are documented examples—not evidence that all suppliers offer equivalent capabilities or that automation is economical for every buyer.