
Zest’s new text-based AI agent helps choose restaurants and book tables.
Wabi, the AI app now positioning itself as a personal agent, says it raised $55 million at a $1.25 billion valuation as investor interest in AI agents accelerates.

Wabi says it raised $55 million at a $1.25 billion valuation. The AI app is now describing itself as a “personal agent,” a positioning shift that places it in the fast-moving race to build consumer-facing AI assistants. The company was founded by Eugenia Kuyda, who previously founded AI companion startup Replika.

Wabi originally focused on letting anyone build apps using prompts. More recently, the startup announced a pivot toward an AI messenger experience, referred to in the article as Wabi 2.0. The company’s new pitch is that Wabi can build the interfaces users need to get tasks done as part of its role as a personal AI agent.
Wabi suggests its mini apps could support everyday planning and personalized tasks. Examples listed include meal planning, tracking local restaurants, planning date nights, creating fitness plans, and designing specialized trivia games. The appeal is practical: instead of only chatting with an AI tool, users may get lightweight interfaces built around the task at hand.
IVP led the new round, with participation from Andreessen Horowitz, Forerunner, Amazon Alexa Fund, and several angel investors. TechCrunch frames the raise against a hot market for AI agents, citing recent funding for Instinct and Tolan as additional signals of investor interest. For readers tracking AI apps, the key takeaway is that Wabi’s valuation reflects confidence in consumer AI agents that move beyond the standard text box.

Zest’s new text-based AI agent helps choose restaurants and book tables.

Meta’s AI fashion app turns wardrobe photos into outfit ideas and trip plans.
![Illustration showing AI agents, depicted by robots, performing digital tasks. [Dom Guzman]](https://news.crunchbase.com/wp-content/uploads/Quarterly-agenticAI.jpg)
A practical outline for scaling AI agents with rules, ownership and risk-based autonomy.

An open smartphone-agent project finds that training across more apps beats simply adding more examples.