Instinct group chats now put an AI agent in the planning

Instinct is taking its personal AI agent out of one-on-one conversations and into the group thread. The startup announced Monday that early access users can now add Instinct to chats with friends, where a shared agent helps the whole group make plans together. The move turns one of messaging's most socially complex spaces into the newest battleground for consumer AI agents.

The idea is simple enough. Group planning usually dissolves into a long back-and-forth over times, places, and who is bringing what. Instinct group chats put a working agent directly in that thread, so the same place where people argue about dinner can also search options, settle on a plan, and get it booked.

The feature works even when the people on the other end have never heard of Instinct. According to TechCrunch's reporting on the announcement, users will be able to add the agent to chats with friends regardless of whether those friends have joined the service. The company described scenarios like planning a trip, snagging event tickets, running a fantasy league, organizing a carpool, or coordinating a shared holiday meal.

That low-friction design matters. Every group chat has at least one person who will not install a new app, and Instinct is betting that the agent can win over the group from inside the thread instead of waiting outside it. The feature is rolling out first to early access users, and users who want it can ask their Instinct agent to put them on the list to try it out, according to TechCrunch's coverage of the launch.

How permission works in a shared thread

A shared agent raises an obvious question: whose information does it see, and who can tell it what to do? Instinct founder Noah Shinn addressed that directly on X. According to Shinn, a user's personal Instinct agent asks for permission before connecting with the group's Instinct agent, and users can choose which groups they trust and revoke that trust at any time.

The group agent is siloed from personal accounts. Shinn explained that the group agent cannot access a user's personal account, and the personal agent always asks the user before sharing information or taking an action on their behalf, reported by TechCrunch. That two-layer design is meant to keep the convenience of a shared assistant from becoming a pipeline that leaks personal data into the chat.

There is also handling for the awkward moment when someone new joins mid-conversation. According to the company's announcement, if new members join the group chat, any pending replies from a user's personal Instinct agent are held back rather than shared with the group. The group agent effectively starts fresh with the new membership, which closes off one of the more subtle ways a shared assistant could expose something a user had not meant to broadcast.

A race for the group thread

Instinct's move puts it a step ahead of its most direct rival in one respect. Meta's Muse, the personal agent the company has been positioning against Instinct, does not yet offer group chat functionality, according to TechCrunch. Meta's older app, Meta AI, can already be added to group chats on WhatsApp, Messenger, Instagram, and Facebook, but Muse is the flagship agent and it is currently a solo experience.

The two companies have been trading feature launches for weeks. Both added the ability to make phone calls recently, and the pace suggests neither expects the consumer agent market to be decided on a single capability. Meanwhile, OpenAI entered the collaborative space last week with ChatGPT Dots, personal agents that can work inside ChatGPT Spaces, shared workspaces where humans and agents collaborate, reported by TechCrunch.

Instinct group chats are a different bet from the workplace-centric approach. Where OpenAI's Spaces aim at teams doing work, Instinct is aiming at friends doing life. The distinction mirrors how messaging actually works: work coordination happens in structured tools, while the logistics of a social life live in threads full of friends who would never join a productivity app.

That positioning also explains the no-signup design. An agent that only works among existing users can never cross the social gap; an agent that can join any conversation has a chance to become the default planner for entire friend groups at once.

Why the group thread is the prize

There is a deeper structural reason the group thread matters to agent builders. A one-on-one agent is a tool. A group agent is a coordination layer, and whoever owns the coordination layer owns the decisions that flow through it: which restaurant gets booked, which tickets get bought, which services the group ends up paying for.

This is the same pattern playing out across consumer AI. Photon AI raised $4.5 million to replace traditional mobile apps with conversational agents spread across messaging platforms, betting that the interface people already use will beat the one they have to download. Hardware makers are making a similar wager: Figure's Hark AI assistant launched with free paid plans, bringing an always-on helper into homes as a device rather than an app.

Instinct has the funding to fight this battle. The startup raised a $1 billion Series C at a $10 billion valuation, with backing from Sequoia Capital, Benchmark Capital, and Coatue, according to TechCrunch's reporting. That war chest is aimed squarely at the thesis that consumers will soon ask AI to take action, not just answer questions, and the group thread is where much of that action gets negotiated.

The risk is trust. A shared agent that books and buys on behalf of a group needs permission handling that is both legible and reliable, because the first time an agent shares the wrong thing in the wrong thread, the whole group leaves. Instinct's siloed-agent design is an attempt to get that right before the stakes scale.

The next test is expansion. The feature is live for early access users now and will open to everyone soon, according to the company's announcement. If groups adopt it, Instinct will have turned the hardest part of messaging — the back-and-forth — into its best distribution channel. If not, it will have learned something expensive about how much agency people actually want in their group chats.