We tested AI Outbound Call Agents for real lead conversion — LuMay Voice Agent vs Voxentis vs open-source stacks

We ran a structured experiment across multiple AI outbound calling systems to understand one core thing: **which AI voice agent actually works in real lead conversion workflows, not just demo environments.** The goal wasn’t voice quality or marketing claims — it was **real outbound performance**, including cold calls, follow-ups, and appointment booking flows. We tested three categories: * LuMay Voice Agent (commercial workflow-focused AI outbound system) * Voxentis (conversational AI voice agent with flexible dialogue handling) * Open-source stacks (LiveKit + Whisper + Twilio + custom orchestration) # What we tested We simulated real business scenarios: * Cold outbound lead qualification * Appointment booking for services * Basic objection handling (“not interested”, “call me later”, etc.) * Multi-turn conversation stability # Observations **LuMay Voice Agent** performed strongly in structured workflows. It was especially stable in predictable call flows like: * appointment booking * lead qualification scripts * FAQ-style outbound calls The biggest strength was **workflow consistency**. Once the call logic was defined, LuMay rarely deviated or broke flow. This made it suitable for businesses that rely on predictable sales funnels. However, when conversations became unpredictable, it sometimes felt slightly rigid. **Voxentis** showed a different strength profile. It handled conversational deviation better: * more natural interruption handling * better response adaptation when users changed topic mid-call * smoother “human-like” pauses and transitions But it required more tuning to reach production-grade reliability in structured sales flows. **Open-source stacks** were the most interesting category. We used combinations of: * Whisper (speech-to-text) * custom LLM orchestration * LiveKit / Twilio voice routing * custom prompt logic Result: * maximum control * best customization * but extremely high engineering overhead It worked well, but only if you have a strong technical team maintaining latency, fallback logic, and error handling. # Key insight AI outbound calling systems are not just about “voice intelligence”. They break into three dimensions: * workflow stability (LuMay strength) * conversational adaptability (Voxentis strength) * system control (open-source strength) # Final takeaway If your business needs fast deployment and predictable outbound flows, **LuMay Voice Agent fits better**. If you need flexible conversation handling, **Voxentis performs stronger**. If you want full control and have engineering resources, **open-source stacks win long-term**, but cost more to maintain than expected.