Lecture Notes: Voice AI in Automation

Introduction

• Presented by Tony, an automation expert from SA Checkout.
• Discussed the creation and use of a voice AI named "Lowballer 9000" to negotiate for a Rolex Daytona watch.
• Emphasizes the potential of voice AI to automate phone-based tasks in the near future.

The Project: Lowballer 9000

• Objective: Automate negotiation with luxury watch dealers for a Rolex Daytona.
• Target: Secondhand luxury watch dealers, not official Rolex outlets.
• Platform Used: Vappy, a flexible Voice API platform.
• AI Name: Alex, designed to gather essential info about the watch (condition, documentation, packaging).
• Prompt Crafting: Developed through testing and mimicking successful negotiation tactics.

Strategies for Effective AI Design

• Conciseness: Responses kept short to avoid overwhelming the dealer.
• Language Level: Uses sixth-grade English to sound more natural and relatable.
• Temperature Setting: Set to influence response creativity; a balance between creativity and predictability.

Tools and Technologies

• Vappy: Allows custom prompts and integration with various services.
• Model Choice: Selection based on current performance and availability (e.g., used DeepSeek during its hype cycle).
• Voice Configuration: Able to select from various persona-like voices (e.g., "New York man").

Integrations and Data Handling

• Tool Calls: Used to interact with existing tools and processes.
• Watch List Database: Tracks all calls and their outcomes.
• Airtable Integration: Records offers and negotiation outcomes.

Challenges in Execution

• Call Connection Issues: Importance of registering phone numbers to avoid calls being flagged as spam.
• FCC Regulation: 'Shake and Stir' enforcement requires phone number registration for calls to be accepted.

Broader Implications and Opportunities

• Sales Automation: Potential applications in real estate for prospecting and lead generation.
• Arbitrage Opportunities: Using AI to identify price gaps in markets (e.g., buying low and selling high).

Business and Startups

• Startup Empire: Tony's involvement in providing tutorials and demos on automation.
• Feedback Collection: Use of voice AI to gather user feedback efficiently, integrated with Slack for real-time updates.

Potential Business Models

• Service Offering: Setting up voice AI systems for businesses to automate feedback collection and data management.
• Partnership Opportunities: Collaborations to expand the use of voice AI across different industries.

Conclusion

• Voice AI offers significant potential for automating phone-based interactions.
• Tony shares insights into creating effective voice AI solutions and encourages exploring arbitrage and sales applications.
• Encourages further exploration into using AI for business automation and startups.