David Petrou, a former Distinguished Software Engineer at Google, has launched Continua, a startup building artificial intelligence agents for group chats. The company has secured $8 million in seed funding to bring what it calls “Social AI” into everyday conversations on platforms such as SMS, iMessage and Discord.
“The simple way to think about it is that we’re bringing the power of LLMs to group chats,” Petrou said.
Petrou spent more than 17 years at Google, contributing to projects including Google Goggles and Google Glass. He left to pursue an idea he believes is better suited to a startup environment: an AI assistant that can participate directly in group conversations.
The funding round was led by GV, formerly Google Ventures, with participation from Bessemer Venture Partners. Notable angel investors joined the round, including Google Chief Scientist Jeff Dean, Intercom co-founder Des Traynor, and many more.
Continua’s AI is designed to act like a participant in a conversation rather than a separate tool. It can follow the flow of messages, remember relevant details, and turn them into checklists, polls, calendar invites or shared documents. It can also perform live web searches without requiring users to leave the chat.
One of the core challenges is making the AI socially aware. Petrou says,” The technology must know when to speak and when to stay silent.” Continua gives users control over the agent’s presence, allowing it to be summoned on demand or asked to step back when needed.
The company points to research showing the scale of the problem it is tackling. Every member of Gen Z and 83 percent of all Americans participate in at least one group chat. Two-thirds of Americans say these chats can be overwhelming. More than half of the workforce reports burnout, with constant coordination playing a role.
Petrou believes AI can help by handling the administrative burden of group communication. The goal is to organise messages and also to understand group dynamics. That means recognising when decisions are forming, keeping track of commitments, and making sure that ideas are not lost in long message threads.
According to a report, the global conversational AI market was valued at over $10 billion in 2024 and is projected to continue expanding as businesses and consumers adopt AI-driven tools for communication and coordination. Several companies are pursuing adjacent concepts, including Meta’s AI-powered group chat features in Messenger, Slack’s Workflow Builder with AI integrations, and emerging startups like Hey Umai and Character.AI, which have experimented with social and multi-user conversational models. This broader industry momentum suggests that tools blending AI with group communication are becoming a significant category within the AI ecosystem.
Continua’s design combines large language models with real-time integrations. It can work across different messaging platforms and connect to services like Gmail and Google Calendar. Users can adjust how actively the AI participates, and they can review and approve actions before they are shared.
GV General Partner Erik Nordlander described Petrou as a visionary engineer with a record of building large-scale, user-facing AI systems. He said, “ Continua’s human-centred approach sets it apart in a crowded field.”
While other AI agents focus on customer service or productivity tools, Continua is being built to work inside existing conversations. Petrou compares it to a helpful meeting participant one that can take notes, highlight action items, follow up afterwards, and occasionally add relevant information without needing to be prompted at every step.
Recent advances in AI infrastructure make this approach possible. Models can now run with lower latency and cost, which is critical for keeping pace with the speed of group chats. The seed funding will go toward expanding the engineering team, adding more platform integrations, and refining the AI’s conversational behaviour.
Early demonstrations show how Continua might work in practice. In one example, the AI joined an iMessage group planning a trip. It pulled out dates and preferences from the chat, created an itinerary, suggested restaurants, and set reminders without anyone leaving the conversation. In another case, a Discord gaming group used it to track tournament scores and share updates in real time.
These tests have shaped features like adjustable verbosity and consent-based actions. Petrou says such design choices are critical for building trust. Social AI, he notes, must avoid the perception of eavesdropping while still being useful. The company has committed to transparent data handling and privacy safeguards. By reducing clutter, clarifying decisions, and keeping track of next steps, Petrou wants Continua to make group conversations more productive and less stressful.
