What to Know Before Building an AI Voice Agent
A conversation with Naada Taiyab, Fractional Head of AI
Before you build an AI voice agent, there’s a lot more to think through than just picking a model or plugging into a voice API. In this episode, Nadaa Taiyab, Fractional Head of AI, shares hard-won lessons from building real-world, HIPAA-compliant voice agents and leading AI initiatives from concept to production. The conversation explores what actually makes a voice agent useful in practice: the underlying model architecture, how to design conversations that feel natural, and how choices around personality, vocal quality, and prompting shape user trust and outcomes.
Nadaa also breaks down why seemingly simple voice workflows, like form filling and structured information capture, can be some of the hardest to get right. From latency and turn-taking to accents, edge cases, and production tradeoffs, this episode is a practical guide for founders, product teams, AI builders and anyone else who want to create voice agents powered by AI that work in the real world, not just in demos.
Key Topics Covered 1. What to know before building an AI voice agent 2. Voice agent architecture and stack design (including models, turn-taking, and core components) 3. Conversation design, personality, and prompt engineering for better user experience 4. Voice quality, accents, and trust in real-world interactions 5. Why “simple” workflows like form filling are hard—and what production teams should expect
Resources to Check Out
About the Guest
Naada Taiyab, Fractional Head of AI
Nadaa Taiyab is a Fractional Head of AI who helps organizations identify high value use cases and lead AI initiatives from concept to production. For nearly a decade, she has built applied AI across national health systems, including clinical and operational workflows and large scale patient data de-identification (U.S. patent). She has also partnered with the Institute for Systems Biology on research to identify patients impacted by adverse social determinants of health. Most recently, she led AI at a healthcare startup, building HIPAA compliant voice agents to automate referral workflows and reduce call center wait times. With experience spanning Goldman Sachs, the World Bank, and health tech startups across three continents, she brings a global perspective and a knack for making complex topics accessible as a speaker and mentor.
About the Host
Marcelo Lewin, Founder @ iBuildWith.ai
Marcelo is the founder of iBuildWith.ai and a Principal Content Architect and Builder at Cigna. He has over 30 years of experience in the tech industry. Having lived through the computer, internet, and mobile revolutions, he is now excited and focused on navigating the AI revolution by helping non-developers build real applications using AI. He founded several startups and held roles at various companies including Toyota, NBC, J.F. Shea, and Walt Disney Imagineering.