A startup that has existed for less than four months is already in talks to raise $100 million at a $1 billion valuation. That alone is worth paying attention to. But what makes Prentis interesting is not the fundraise, it’s the specific bet the company is making: that automating routine office work will become a bigger AI opportunity than writing code.
According to TechCrunch, Prentis was founded in April by Ritankar Das alongside LinkedIn co-founder Reid Hoffman and Zynga founder Mark Pincus. The lab is training models to watch how office workers move through documents, software, and systems, then building agents that can replicate and automate those workflows. Think insurance claims processing, customs duty refund exceptions, supply chain paperwork. The kind of work that is genuinely time-consuming, deeply repetitive, and currently requires a human to click through a dozen screens.
The company says it has already signed contracts worth up to $50 million with customers in healthcare administration, manufacturing, and retail. Its own investor materials project an annualized run rate of $75 million by Q3 of this year, though Prentis notes those figures are based on a 20% share of realized savings, not booked revenue, and are performance-dependent. So read that number with appropriate caution.
The technical claim at the center of Prentis’s pitch is that its Hive-32B model outperforms OpenAI’s GPT-5.4 and Anthropic’s Claude Opus 4.6 on two computer-use benchmarks: WindowsAgentArena, which tests end-to-end task completion inside real Windows applications, and ScreenSpot-v2, which measures a model’s ability to identify the correct on-screen element. The company also claims roughly ten times lower cost per task than frontier APIs, arguing that a smaller, cheaper model is actually better suited to repetitive enterprise workflows than a general-purpose frontier model. TechCrunch has not independently verified these results.
But the market Prentis is entering is already crowded. Anthropic, OpenAI, and Mira Murati’s Thinking Machines Lab are all working on computer-use agents. Anthropic moved aggressively earlier this year by acquiring Seattle-based Vercept, a startup focused specifically on this problem, shutting down its product and bringing the founders in-house. That acquisition signals how seriously the frontier labs are taking this category.
Prentis is led by Das, who is 31 and runs Titan, a self-funded holding company that has previously built healthcare AI companies including Tala Health and Forta Health. Hoffman and Pincus are co-founders but operating in a more advisory capacity. Hoffman is currently focused on Manas AI, his drug-discovery venture, after stepping down from Microsoft’s board. Pincus runs Reinvent Capital.
The team Prentis has assembled includes researchers from OpenAI, Google DeepMind, Meta, Tencent, and Alibaba, which suggests it is not just a founder story dressed up as a lab. Still, the real test will be whether the benchmark performance holds up in production, and whether enterprise customers keep renewing once the initial contracts are fulfilled.
