Anthropic Acquires Agentic AI Startup Vercept
Anthropic has acquired Vercept, a Seattle-based startup that builds agentic AI tools to automate multi-step tasks within software applications. Vercept's product will be discontinued as its co-founders join Anthropic to integrate the technology into the Claude AI model. The acquisition aims to bolster Claude’s ability to operate inside real-world enterprise workflows.
Vercept was founded in 2024 by a team of former researchers from the Allen Institute for AI (AI2), including CEO Kiana Ehsani and co-founders Matt Deitke, Ross Girshick, Luca Weihs, and former AI2 CEO Oren Etzioni. The startup raised over $50 million, including a $16 million seed round in January 2025 with backing from prominent angel investors like former Google CEO Eric Schmidt and Google DeepMind chief scientist Jeff Dean. Vercept's flagship product, Vy, was a cross-platform AI agent designed to control a computer through natural language, automating tasks by observing the screen and manipulating the mouse and keyboard. The company claimed its proprietary VyUI model outperformed those from OpenAI, Google, and Anthropic on user interface grounding benchmarks. This acquisition is Anthropic's second, following the purchase of the coding toolkit Bun in late 2025, signaling a strategy of acquiring small, technically elite teams to accelerate specific capabilities. The move intensifies the "agentic AI" arms race, where AI transitions from a passive assistant to an active operator within enterprise software. This market is projected to grow to between $12 and $15 billion in 2026, with M&A activity surging; 16 acquisitions occurred in the sector in 2025, up from 6 in 2024. Major players like Google DeepMind and Microsoft are heavily invested in their own agent initiatives, such as Project Astra and Copilot, respectively. This push towards agentic AI is creating a pincer movement in the hardware market, driving demand for both specialized, custom-built silicon in data centers and more powerful on-device AI chips. Hyperscalers like Google and Microsoft are developing their own ASICs, such as Google's Ironwood TPU and Microsoft's Maia, to reduce inference costs by a reported 40-60% compared to traditional GPUs for predictable, high-volume workloads like autonomous agents. For go-to-market teams, the rise of agentic AI is rewriting the sales playbook. Venture capitalists are increasingly backing startups that leverage AI for GTM, with some investors stating they are less inclined to fund companies not using AI tools. AI is being used to automate everything from lead scoring and competitive intelligence to crafting personalized outreach, enabling leaner teams to achieve higher revenue per headcount.
