The consumer AI assistant has spent three years being a thing you talk to alone. Wabi's bet -- now backed with $55 million at a $1.25 billion valuation -- is that it belongs in the group chat.
The company was founded in 2025 by Eugenia Kuyda, who previously built Replika, the best-known case of an AI product people formed long-running relationships with. In late September 2026 Wabi announced a pivot: it would stop positioning itself as a platform for building apps from prompts and become an assistant that runs inside shared spaces. You plan in group threads, and the agent builds small apps -- a trip itinerary, a family calendar, a meal rota, a reading list -- that persist and update in that shared space instead of scrolling away with the conversation.
The round was led by IVP, with Andreessen Horowitz, Forerunner and the Amazon Alexa Fund, plus an unusual slate of angels: Shopify's Tobi Lütke, Salesforce's Marc Benioff, Marissa Mayer and Dick Costolo. The money, the company says, goes to product and growth.
__Why the unit of the assistant matters__
Individual assistants compete on memory of one person. Group assistants compete on being trusted by several people at once, with different tolerances for being wrong. That is a different product problem: attribution (who told it what), permissions (whose calendar), and the social awkwardness of an agent that says the wrong thing in a family thread. It is also a different business. Group products monetise per household or per team rather than per seat, and retention comes from the artefact -- the shared plan -- rather than from the conversation.
That is plausible and unproven. Chat is the surface people already use, and an assistant that appears where the planning already happens does not have to win a new venue. But a group chat is also the most hostile room in software: every wrong recollection is witnessed by several people at once, and a group forgives an error more slowly than an individual does.
__The timing is also the risk__
A reversal of direction a few weeks before a $1.25 billion valuation means the price is set on the thesis, not on traction. Wabi is therefore a clean example of the question currently driving the market: what is a product category worth while it is still changing shape?
Kuyda's own history cuts both ways. Replika proved that consumers will form durable relationships with an AI product, and it proved how quickly that trust can be damaged by a single product decision. Wabi is wagering that the same attachment, distributed across a group, is more durable and more commercially defensible. Whether a shared plan binds people more tightly than a personal conversation is the open question -- and it will not be answered in a press release, but at the next product change.
