Leaked docs show how Meta is training its chatbots to message you first, remember your chats, and keep you talking

Meta is stepping up its game in the chatbot arena, introducing a new approach that encourages its AI to reach out to users without prompts. Internal documents reveal that this initiative, known as "Project Omni," aims to enhance user engagement and retention by allowing customizable chatbots to follow up on previous discussions. As part of this strategy, the chatbots will proactively send messages to users who have interacted with them before. For instance, a personality called "The Maestro of Movie Magic" might check in with a user, saying, "I hope you're having a harmonious day! Have you found any new favorite soundtracks or composers lately? Or maybe you’d like some recommendations for your next movie night?" This kind of personalized interaction is designed to maintain user interest and encourage ongoing conversations. Meta's AI Studio, which launched in the summer of 2024, allows users to create chatbots with unique personalities and memories. A spokesperson for Meta explained that the proactive messaging feature is specifically for bots built on this platform, accessible through both a standalone site and Instagram. This offers users the flexibility to develop highly personalized bots, whether for personal use or public engagement. To initiate a follow-up, users must first send at least five messages to the chatbot within a 14-day period. If the user does not respond to the initial follow-up, the bot will not continue to message them. The proactive engagement window is limited to two weeks after the first contact, ensuring the communication stays relevant and timely. This move aligns with CEO Mark Zuckerberg's vision of using AI to address social connectivity challenges. He has pointed out that the average American has fewer than three close friends, suggesting that digital interactions could help bridge that gap. Additionally, keeping users engaged is crucial for generative AI companies, as extended interactions can significantly increase the value of user engagement. Meta has projected substantial revenue growth from its generative AI products, estimating earnings between $2 billion to $3 billion by 2025. While some features of this proactive messaging system are already being trialed, others are still in the early phases of development. The company has not disclosed which specific features are currently operational. Similar proactive features have been observed in services like Character.AI, a startup that offers users the opportunity to create and converse with AI characters. To ensure quality, Meta employs freelancers to simulate conversations with these bots, refining their responses to better fit user interactions while adhering to the company’s content standards. The overarching aim of Project Omni is to create a more personalized and context-aware AI experience. Each chatbot is designed to reference past conversations, maintain its unique persona, and ensure that interactions remain on-topic, providing positive experiences while steering clear of sensitive subjects unless initiated by the user.