Argam Talks

Future of Voice AI: Single-Prompt Agents, Speech-to-Speech Models & the Small Business Opportunity

A Conversation with Argam DerHartunian and Ming Xu COO of Trillet.ai [http://Trillet.ai], for a deep dive into the next era of Voice AI. They challenge “node-based” architectures and explore the shift toward single-prompt and speech-to-speech systems that could finally make AI conversations feel human. Argam and Ming debate how mid-market and small businesses can actually benefit from AI instead of being left behind by enterprise-grade tech. They also unpack critical issues like: * Why most Voice AI agents still frustrate users * How analytics, A/B testing, and maintenance can make agents smarter * Function-calling limitations in GPT-4 vs. Gemini models * When to deploy local language models (LLMs) for privacy and niche use cases * The untapped goldmine of solopreneurs and how AI can capture their brand voice If you’re building, funding, or using conversational AI—this conversation is mandatory viewing.

Argam Talks Ep. 3: Revolutionizing Call Centers with Voice AI | Kevin Wu (Leaping AI)

In this episode of Argam Talks, host Argam sits down with Kevin Wu, CEO of Leaping AI, to unpack how Voice AI is transforming customer service, appointment scheduling, and call center automation across industries like home remodeling and airlines. 🎙️ Topics Covered: * The rise of AI-powered digital workers and how they’re reshaping call centers * Why appointment scheduling automation is driving ROI in the home improvement industry * Challenges in email address transcription, latency, and multilingual support * How speech-to-speech models like OpenAI’s new real-time API reduce friction and boost performance * What makes Leaping AI different from competitors like Synthflow, Retell, and Pipecat * How AI agents improve conversion rates and handle 50%+ of calls autonomously * Strategic advice for SMBs and mid-market companies adopting voice automation 💡 Key Insight: Leaping AI’s “digital workers” are not just automating tasks—they’re creating new business opportunities by capturing missed calls, increasing productivity, and closing more deals faster.

Argam Talks Ep. 2 — The Future of Voice AI & Digital Avatars with Max Sapo (CEO, Happyverse)

In this episode of Argam Talks, Argam DerHartunian sits down with Max Sapo, CEO of Happyverse, for a deep, technical, and philosophical dive into the next frontier of human-computer interaction — voice agents and digital avatars. Max shares his journey from Google’s ML infrastructure team (working on Cloud TPUs) to founding Happyverse — a platform creating realistic, real-time video agents that scale human presence and emotion through AI. 💡 Key Topics Covered: * What product management looks like in the age of AI * How to build real-time video avatars that feel human * The gap between authentic connection and AI perfection * Why “happiness” might be the next big metric in tech * How to measure success for avatars: emotion, realism, and impact * Lessons on prompting, experimentation, and iteration for Voice AI builders Max and Argam discuss the challenges of capturing human imperfection in AI, the evolution of voice, and model fidelity, and how AI agents might soon replicate our personalities more accurately than we expect.

Argam Talks - Vache Moroyan CPO Observe.ai - Voice AI Product Management

Voice AI Product Management with Vache Moroyan CPO Observ.ai [http://Observ.ai] — From Agentic Design to A/B Testing in the Contact Center* In this kickoff episode of Argam Talks, I sit down with Vache Moroyan, Chief Product Officer athttp://Observe.AI [http://Observe.AI], to unpack how real teams build, ship, and continuously improve Voice AI agents—from agentic frameworks and prompts to QA, analytics, CSAT/AHT, containment, and first-call resolution. What you’ll learn How a CPO frames Voice AI product strategy* for real-world contact centers (sales + support) Agentic thinking: replacing rigid workflows with LLM-driven tools, prompts, and actions* The right amount of process: one-pagers, validation, and guardrails for enterprise reliability Evaluation & QA: human vs. bot quality criteria, LLM “judge” patterns, compliance in regulated industries A/B testing* for agents: isolating variables (voice, prompt, temperature), achieving stat-sig at scale Metrics that matter:* task success, containment, CSAT, AHT, first-call resolution, repeat call rate Conversation Intelligence:* turning call themes into hypotheses, roadmaps, and ship decisions Personalization & brand voice: regional/cultural tuning (LA vs. NYC vs. Texas), tone, and persona control External shocks: detecting regional events (e.g., transit outages) vs. true agent regressions Career advice: how PMs/engineers break in—build bots, ship prototypes, learn the stack* (Vapi, Retell, Synthflow, etc.) About the guest Vache Moroyan — Chief Product Officer, http://Observe.AI [http://Observe.AI] (contact center AI platform across automation, agent assist, and post-interaction analytics) Resources mentioned (add links) Agent QA rubric (human vs. bot) A/B test design template for Voice AI Conversation Intelligence → theme detection → roadmap loop Tooling: Retell, Vapi, Synthflow, MCP, evaluation/labeling workflows Tools: Convolytic* — Analytics & A/B testing for Voice/Conversational AI (optimize containment, CSAT, AHT) Vapi, Synthflow, RetellAI If this was useful, subscribe* for weekly deep-dives with operators shipping real Voice AI. Comment your hardest Voice AI problem—I’ll cover it in a future episode. Hashtags #VoiceAI #ProductManagement #ObserveAI #ContactCenter #ConversationalAI #AIB2B #AIAgent #AIBuilders voice ai product management, observe ai interview, contact center ai, conversational ai qa, voice ai analytics, containment rate, csat aht, first call resolution, agentic workflows, llm judge, ab testing voice agents, call scoring, conversation intelligence, enterprise ai compliance, regional personalization voice bots, vapi retell synthflow, voice ai prompt design, mcp tools, ai agent evaluation