Voice AI is leaving the demo phase and going enterprise. Thoughtly CTO Alex Casella dug into what it really takes to run production voice agents during a webinar with Gladia CEO Jean-Louis Quéguiner: 1. ⚡ Latency defines realism — optimize every hop (STT → LLM → TTS → VoIP) and benchmark constantly. 2. 🧭 Evals are your compass — build regression tests early so you can swap models without surprises. 3. 🧩 Orchestrate, don’t over-host — focus your effort where it creates customer value. 4. 🔒 Compliance & infra matter — SOC 2/HIPAA readiness and reliable, global infra win enterprise trust. 5. 🌍 Multilingual reach — quality, speed, and language coverage beat brand names. Book a demo and turn every lead into a live conversation with Gladia TTS + Thoughtly orchestration: 👉 thought.ly/demo #VoiceAI #GTM #ContactCenter #AIAgents #SpeechTech #TTS #STT #LLM #Latency #Compliance #EnterpriseAI #Thoughtly #Gladia
Voice AI is entering its Mature Era, moving from experiments to enterprise-grade systems. In this conversation between Alex Casella (CTO @ Thoughtly) and our CEO, Jean-Louis Quéguiner, we covered what it really takes to build and scale production-ready voice agents: ⚙️ latency and orchestration 🧠 evals and benchmarking 🛡️ compliance and infra strategy 🚀 and how to find product-market fit in a fast-moving market In the bite-sized preview below, Alex and JL also touch on one of the biggest strategic questions for any voice AI builder: Should you self-host your models, or rely on specialized vendors? The answer is not always clear-cut, but for Alex, when it comes to voice agent builders, one thing is certain: we’ve reached the point where the market moves faster than any one startup can keep up by hosting its own models. Read the full recap or watch the replay → https://lnkd.in/eY8n33xy