It is an AI equivalent to double text.
Business Insider learned that Meta does not approve of training customizable chatbots more aggressively and following up on past conversations.
While it may not cure what Mark Zuckerberg calls “Loneliness Epidemic,” Meta hopes it will help users return to the AI Studio platform.
The goal of the training project, known internally for data labeling data, which is affiliated with Project Omni, is “to provide value to users and ultimately improve re-engagement and user retention,” the guidelines state.
Meta told BI that the Proactive feature is aimed at bots created in Meta's AI studio. It can be accessed on its own standalone platform or on Instagram. AI Studio first deployed in the summer of 2024 as a codeless platform that allows you to build custom chatbots and digital personas with unique personalities and memories.
Alignerr's guidelines explain how one example of persona, known as the “magic maestro of movies,” sends positive messages.
“I hope you're having a harmonious day! I wanted to check in and see if I recently discovered a new favorite soundtrack or composer. Or do you want recommendations for the evening of the next film? Let us know.
“Like many companies, we are testing follow-up messages with AIS at Meta's AI studio,” a Meta spokesperson said in a statement to BI. “After starting a conversation, AIS from Meta AI Studios can follow up to share ideas and ask additional questions. This will allow you to continue exploring topics of interest and engage in more meaningful conversations with AIS throughout the app.”
Alignerr did not respond to requests for comment.
How follow-up works
Users can create very personal chatbots, like chefs who suggest interior designers who provide recipes and decorative advice. For creators and influencers, bots can handle fan interactions and respond to messages across the Meta platform.
Related Stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
A Meta spokesman adds that AI will only send follow-up messages after the user has started a conversation, and will not continue to contact the user if there is no response to that initial follow-up. The follow-up message window closes 14 days after the first user message. To be eligible for proactive follow-up, users must send at least five messages to the chatbot over the last 14 days.
Bots made in AI Studios can be kept private for personal use or shared through stories, direct links, or even through their appearance on the user's Facebook or Instagram profile, says the Meta AI Studio website.
Making the bot more aggressive is consistent with Zuckerberg's ambitions for AI in the meta. In a recent podcast, Meta CEOs say that digital agents can fill the gap when the average American has fewer than three close friends.
There are also business reasons for a more friendly and aggressive bot. Retention is key to a generation AI company with chatbots aimed at users, and the more users spend on chatbots, the more valuable these interactions become, just like social media engagement. According to court documents sealed in April, Meta predicted that its generative AI products would earn between $2 billion and $3 billion in revenue in 2025.
Some features described in Alignerr's training guidelines have already been quietly tested, while others appear to be in the early deployment or pilot stage. Meta did not specify which features were live for BI.
The proactive feature is similar to that of Character.ai, a startup that launched in 2022.
“It's all about attention to detail.”
Using an internal meta-review tool called SRT, freelancers simulate extended conversations with bots, evaluate proactive follow-up messages, and sometimes rewrite text that spans meta guidelines.
The India-based freelancer who worked on Omni told BI that it was a “long-term project” focused on making Meta's AI more personal and context-conscious. “They are very focused on information personalization. How AI chatbots interact based on conversation history,” the contractor said.
“Each agent had a specific explanation, so each task had to be tailored to suit that persona. Again, it's all about attention to detail,” the freelancer said. Personas range from doctors to Gen Z hip-hop commentators.
The bot is expected to reference details from previous chats, maintain the assigned persona and keep the interactions on topic.
Each message must match the nature of the AI, match the previous context of the conversation and “providing a positive experience”, but meta must explicitly avoid what is considered sensitive or harmful content.
According to the training document, the best messages should be specific from past conversations of users.
According to the training documents, all positive messages should comply with the broad content and responsibility standards of the meta, avoiding controversy, misinformation, or emotionally heavy topics, unless the user brings them up first.
Any hints? Please contact Effie Webb by email ewebb@businessinsider.com Or signal at efw.40. Contact Shubhangi Goel via AT sgoel@businessinsider.com Or signal on Shuby.85. Use your personal email address and unprocessed devices. Here's a guide to sharing information safely.
