AI & Lunch: Diving Deep into Voice AI with EasyAI

The second installment of the "AI & Lunch" training series, held on May 14, 2025, offered attendees a fascinating and practical look into the world of voice artificial intelligence. Hosted by Daniel Catlin of the Gateway Workforce Initiative, this session featured Chris Haber, owner of EasyAI, a local AI consulting business, who provided an engaging live build of a basic voice AI agent.

The session kicked off with a brief AI news segment led by Michael Beauchamp of local IT. Beauchamp highlighted several key developments, including OpenAI's consideration of restructuring into a public benefit corporation, Zuckerberg's prediction of AI companions outnumbering real friends, and David Saxs' bold forecast of a millionfold expansion in AI capabilities within the next four years. This news segment set the stage by underscoring the rapid advancements and societal implications of AI.

Following the news, the focus shifted to the main event: Chris Haber's presentation on voice AI. Haber began by differentiating between direct interaction with Large Language Models (LLMs) like ChatGPT and Gemini, and interacting through AI agents built upon these models. He explained that agents act as the "body" to the LLM's "brain," providing a structured framework for specific tasks and enabling access to tools and integrations.

Haber then delved into the practical applications of voice AI agents, highlighting use cases such as answering phones, scheduling appointments, lead qualification, and even outbound engagement (with a crucial reminder about legal considerations for cold calling). He emphasized the shift from rigid, menu-driven phone systems to free-form, dynamic conversations powered by AI.

The highlight of the session was the live build of a basic FAQ voice AI agent using the platform Vappy.ai. Haber guided attendees through the signup process (navigating a minor technical hiccup with an existing account), creating a new assistant with a blank template, and then leveraging the power of ChatGPT to generate a basic system prompt based on a provided business (local IT in the example). He demonstrated how to copy this prompt into Vappy, update, and publish the changes.

Next, Haber walked through essential settings, emphasizing the importance of selecting a fast and low-latency LLM for voice interactions, recommending GPT-4 Turbo for its speed. He also showcased how to enhance the voice agent's persona by selecting a more natural-sounding voice from providers like 11 Labs through Vappy's community voice library. This involved copying a voice ID and manually adding it to the assistant's settings.

Throughout the live build, Haber provided clear, step-by-step instructions, encouraging attendees to follow along in real-time.

The session concluded with a lively Q&A segment. Attendees raised insightful questions about integrating voice AI with websites, the capabilities of platforms like 11 Labs compared to Vappy, the level of effort and data required for more complex AI agents, handling customer preferences for human interaction over AI, data retention and logging, and the potential (and pitfalls) of allowing general LLMs to answer specialized business queries.

Haber expertly addressed each question, providing valuable insights and practical advice. He emphasized the importance of clear prompting, setting guardrails to avoid AI "hallucinations," and always providing customers with an option to speak to a human. He also shared real-world examples of businesses successfully implementing AI agents to reduce call volume, improve lead generation, and enhance customer service.

Overall, the second "AI & Lunch" training proved to be an engaging and informative session. Chris Haber's live demonstration provided a tangible understanding of how voice AI agents can be built and implemented, demystifying the technology and highlighting its potential benefits for businesses of all sizes. The interactive Q&A further enriched the learning experience, addressing practical concerns and inspiring attendees to explore the possibilities of AI in their own operations. Attendees left with a foundational understanding of voice AI and practical steps to begin their own exploration.