The video features Shivani Poddar, CEO of Wajro AI, discussing their AI assistant Four, which uniquely combines AI with human experts to reliably complete complex tasks while prioritizing user privacy through strict data controls. This hybrid approach enables Four to handle diverse real-world tasks effectively, earning strong user trust and feedback, with plans to keep the consumer version free and focus monetization on business applications.
The video features an interview with Shivani Poddar, CEO of Wajro AI, discussing their recently launched personal consumer AI assistant called Four. Since its launch about two and a half weeks ago, Four has rapidly grown its user base and is designed to be a trustworthy AI agent that can complete complex tasks across various ecosystems without compromising user privacy. Unlike many AI assistants, Four incorporates a unique human-in-the-loop system, where trained human experts step in to assist with tasks that the AI cannot fully handle, ensuring reliability and thorough completion of user requests.
Shivani emphasizes that the future of AI lies in collaboration between AI and humans rather than AI replacing humans. Their approach, called subtask delegation, allows the AI to identify when it needs help and bring in human experts who are trained and vetted to handle administrative tasks, negotiations, and even specialized roles like stylists. This hybrid model enables Four to tackle increasingly complex tasks, from negotiating bills to managing business operations, providing a level of service that purely AI-driven agents struggle to achieve.
Privacy and trust are central to Four’s design. Unlike many AI systems that have open access to user data, Four employs audience-scoped access control, ensuring that personal information is only shared with explicit user consent. Shivani, drawing on her experience in AI safety at Gemini, highlights that Four never uses private user data for training its models. Instead, it learns from anonymized, expert-generated examples, prioritizing user security and data protection while maintaining high performance.
The interview also delves into Shivani’s personal journey, from studying natural language generation at Carnegie Mellon and working on early conversational agents like Alexa, to her roles at Facebook and Gemini, before founding Wajro AI. Her vision is to create autonomous, proactive AI agents that can handle real-world tasks seamlessly. The company operates with a high level of trust in Four, using it internally for critical functions such as recruiting, where Four manages candidate sourcing, testing, and scheduling interviews autonomously.
User feedback has been overwhelmingly positive, with diverse user groups benefiting from Four, including families, small businesses, and individuals managing complex tasks like Medicaid paperwork or job applications. The AI assistant has even been trusted to negotiate significant discounts on purchases and handle sensitive tasks like interviewing nannies. Shivani envisions Four remaining free for consumer use, with monetization focused on business applications, and firmly rejects the idea of incorporating ads to maintain user trust. Overall, Four represents a pioneering step in blending AI capabilities with human expertise to deliver reliable, privacy-conscious personal assistance.