StarZero All shows Try the podcast clipper →

May 2, 2025 · 23m · startup-ideas

My Voice AI Agent Negotiated 800+ Business Deals in 1 Day (FULL Tutorial)

Tony Ge · 13m spoken Greg Isenberg · 6m spoken Alex (Baker Time) · 3s spoken Watch Dealer (Robert Jewelers) · 2s spoken Watch Dealer (Prestige Town Stahl) · 2s spoken Ed (Jewels in Time) · 2s spoken Watch Dealer (Jewels in Time) · 1s spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

Host Greg Isenberg and automation expert Tony break down how an autonomous voice AI agent negotiated over 800 luxury watch deals, detailing the technical architecture, conversational prompting, backend integrations, and commercial arbitrage opportunities of voice AI.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Greg holds 30.3% of the talking time here. How this is scored →

Greg as informed peer 2.8 Guest teaching 4.4 Guest disagreement 0.0 Greg pushing back 0.6
05100:0010:0020:001:28–6:01 · Greg as informed peer 2/10 Prompt Engineering and Conversational Design in Vapi Tony provides a detailed walkthrough of conversational prompt engineering in Vapi, emphasizing concise responses and a sixth-grade English level to keep call durations high. Greg primarily asks clarifying questions and affirms Tony's logic.6:01–8:09 · Greg as informed peer 1/10 Language Model Selection and System Latency Pitfalls Tony explains the real-world operational hazards of using hypetrain LLMs like DeepSeek during peak demand versus more reliable providers like Gemini Flash. Greg asks a basic selection question followed by a mid-roll community plug.8:09–12:45 · Greg as informed peer 2/10 Voice Synthesis, LLM Temperature, and Persona Configuration Tony breaks down LLM temperature settings and how tool calling parses 800+ raw phone transcripts into structured Airtable offer data. Greg participates with light banter regarding voice persona accents.12:47–15:45 · Greg as informed peer 2/10 Integrating Lindy for Dynamic Unstructured Data Extraction Tony demonstrates how Lindy performs zero-configuration column inference on unstructured data and explains STIR/SHAKEN telecommunication compliance. Greg nudges Tony back onto tooling when the guest briefly digresses.15:45–19:37 · Greg as informed peer 7/10 Market Arbitrage Models and Future Voice AI Applications Greg steps into an expert strategist role, articulating a generalized business framework for using voice agents to exploit market information asymmetry and arbitrage. Tony readily agrees and validates Greg's analysis.1:28–6:01 · Guest teaching 5/10 Prompt Engineering and Conversational Design in Vapi Tony provides a detailed walkthrough of conversational prompt engineering in Vapi, emphasizing concise responses and a sixth-grade English level to keep call durations high. Greg primarily asks clarifying questions and affirms Tony's logic.6:01–8:09 · Guest teaching 4/10 Language Model Selection and System Latency Pitfalls Tony explains the real-world operational hazards of using hypetrain LLMs like DeepSeek during peak demand versus more reliable providers like Gemini Flash. Greg asks a basic selection question followed by a mid-roll community plug.8:09–12:45 · Guest teaching 5/10 Voice Synthesis, LLM Temperature, and Persona Configuration Tony breaks down LLM temperature settings and how tool calling parses 800+ raw phone transcripts into structured Airtable offer data. Greg participates with light banter regarding voice persona accents.12:47–15:45 · Guest teaching 6/10 Integrating Lindy for Dynamic Unstructured Data Extraction Tony demonstrates how Lindy performs zero-configuration column inference on unstructured data and explains STIR/SHAKEN telecommunication compliance. Greg nudges Tony back onto tooling when the guest briefly digresses.15:45–19:37 · Guest teaching 2/10 Market Arbitrage Models and Future Voice AI Applications Greg steps into an expert strategist role, articulating a generalized business framework for using voice agents to exploit market information asymmetry and arbitrage. Tony readily agrees and validates Greg's analysis.1:28–6:01 · Guest disagreement 0/10 Prompt Engineering and Conversational Design in Vapi Tony provides a detailed walkthrough of conversational prompt engineering in Vapi, emphasizing concise responses and a sixth-grade English level to keep call durations high. Greg primarily asks clarifying questions and affirms Tony's logic.6:01–8:09 · Guest disagreement 0/10 Language Model Selection and System Latency Pitfalls Tony explains the real-world operational hazards of using hypetrain LLMs like DeepSeek during peak demand versus more reliable providers like Gemini Flash. Greg asks a basic selection question followed by a mid-roll community plug.8:09–12:45 · Guest disagreement 0/10 Voice Synthesis, LLM Temperature, and Persona Configuration Tony breaks down LLM temperature settings and how tool calling parses 800+ raw phone transcripts into structured Airtable offer data. Greg participates with light banter regarding voice persona accents.12:47–15:45 · Guest disagreement 0/10 Integrating Lindy for Dynamic Unstructured Data Extraction Tony demonstrates how Lindy performs zero-configuration column inference on unstructured data and explains STIR/SHAKEN telecommunication compliance. Greg nudges Tony back onto tooling when the guest briefly digresses.15:45–19:37 · Guest disagreement 0/10 Market Arbitrage Models and Future Voice AI Applications Greg steps into an expert strategist role, articulating a generalized business framework for using voice agents to exploit market information asymmetry and arbitrage. Tony readily agrees and validates Greg's analysis.1:28–6:01 · Greg pushing back 0/10 Prompt Engineering and Conversational Design in Vapi Tony provides a detailed walkthrough of conversational prompt engineering in Vapi, emphasizing concise responses and a sixth-grade English level to keep call durations high. Greg primarily asks clarifying questions and affirms Tony's logic.6:01–8:09 · Greg pushing back 0/10 Language Model Selection and System Latency Pitfalls Tony explains the real-world operational hazards of using hypetrain LLMs like DeepSeek during peak demand versus more reliable providers like Gemini Flash. Greg asks a basic selection question followed by a mid-roll community plug.8:09–12:45 · Greg pushing back 0/10 Voice Synthesis, LLM Temperature, and Persona Configuration Tony breaks down LLM temperature settings and how tool calling parses 800+ raw phone transcripts into structured Airtable offer data. Greg participates with light banter regarding voice persona accents.12:47–15:45 · Greg pushing back 1/10 Integrating Lindy for Dynamic Unstructured Data Extraction Tony demonstrates how Lindy performs zero-configuration column inference on unstructured data and explains STIR/SHAKEN telecommunication compliance. Greg nudges Tony back onto tooling when the guest briefly digresses.15:45–19:37 · Greg pushing back 2/10 Market Arbitrage Models and Future Voice AI Applications Greg steps into an expert strategist role, articulating a generalized business framework for using voice agents to exploit market information asymmetry and arbitrage. Tony readily agrees and validates Greg's analysis.

speaking balance: gold is Greg, purple is the guest (3 minute bins)

0:00 · Greg 13% · guest 87%0:00 · Greg 13% · guest 87%3:00 · Greg 11.2% · guest 88.8%3:00 · Greg 11.2% · guest 88.8%6:00 · Greg 45.4% · guest 54.6%6:00 · Greg 45.4% · guest 54.6%9:00 · Greg 9.9% · guest 90.1%9:00 · Greg 9.9% · guest 90.1%12:00 · Greg 15.6% · guest 84.4%12:00 · Greg 15.6% · guest 84.4%15:00 · Greg 45.6% · guest 54.4%15:00 · Greg 45.6% · guest 54.4%18:00 · Greg 39.8% · guest 60.2%18:00 · Greg 39.8% · guest 60.2%21:00 · Greg 77.3% · guest 22.7%21:00 · Greg 77.3% · guest 22.7%
Sharpest disagreement ▶ 9:47 Tongue-in-cheek voice persona disclaimer

In an entirely cooperative episode, Tony playfully parries Greg's Sopranos New Jersey comment by labeling it alleged allegations.

 Hardest push from Greg ▶ 14:44 Steering focus back to technical tools

Greg immediately redirects Tony after a quip about watch dealer attitudes to ensure the conversation remains focused on tooling requirements.

 Biggest teaching moment ▶ 14:49 STIR/SHAKEN telephony compliance lesson

Tony educates the host on FCC telephony authentication protocols that cause unregistered automated outbound calls to drop straight to voicemail.

 Greg holds their own ▶ 17:12 Information asymmetry arbitrage framework

Greg demonstrates strong business acumen by abstracting the watch demonstration into a broader business thesis around data arbitrage.
the scores for every segment, with the reasoning behind each
ChapterTopicGreg as informed peerGuest teachingGuest disagreementGreg pushing backWhy
Prompt Engineering and Conversational Design in Vapi 2500 Tony provides a detailed walkthrough of conversational prompt engineering in Vapi, emphasizing concise responses and a sixth-grade English level to keep call durations high. Greg primarily asks clarifying questions and affirms Tony's logic.
Language Model Selection and System Latency Pitfalls 1400 Tony explains the real-world operational hazards of using hypetrain LLMs like DeepSeek during peak demand versus more reliable providers like Gemini Flash. Greg asks a basic selection question followed by a mid-roll community plug.
Voice Synthesis, LLM Temperature, and Persona Configuration 2500 Tony breaks down LLM temperature settings and how tool calling parses 800+ raw phone transcripts into structured Airtable offer data. Greg participates with light banter regarding voice persona accents.
Integrating Lindy for Dynamic Unstructured Data Extraction 2601 Tony demonstrates how Lindy performs zero-configuration column inference on unstructured data and explains STIR/SHAKEN telecommunication compliance. Greg nudges Tony back onto tooling when the guest briefly digresses.
Market Arbitrage Models and Future Voice AI Applications 7202 Greg steps into an expert strategist role, articulating a generalized business framework for using voice agents to exploit market information asymmetry and arbitrage. Tony readily agrees and validates Greg's analysis.

Statements from this episode (1)

Opinion
Isenberg: Every Company Would Pay for Voice AI Feedback Collection
“Go do that for other companies. Every company would pay for that.”
Greg Isenberg May 2, 2025 ▶ 22:07
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 300 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more