“Our AI agent researches prospects, drafts personalized messages, and books meetings — autonomously.”Claim evidence: source page
What it actually means
Autonomous here means LLM-driven sequence orchestration with human-in-the-loop guardrails enabled by default. The page doesn't show what happens when those guardrails are turned off.
How to test it
Run a parallel pilot: one team uses the agent, one team uses the existing SDR workflow. Compare AE-accepted meetings, reply quality scores, and unsubscribe complaints over four weeks.
▶4 hidden assumptions
- Research depth is high enough to clear AE bullshit detectors.
- Personalization survives at sequence-level scale, not just demo-level.
- Booked meetings are AE-accepted, not just calendar-blocked.
- Reply handling escalates the right exceptions to a human.
Roast: Autonomy is the marketing word. The release notes still say 'beta, human review recommended.'