Prentis AI Lab Seeks $100M at $1B Valuation to Automate Office Workflows
The computer-use startup, backed by Reid Hoffman and Mark Pincus, has signed up to $50M in customer contracts while still in fundraising talks.
A new AI research lab called Prentis is in discussions to raise $100 million at a $1 billion valuation, according to two people familiar with the talks. The round has not yet closed, and no Form D has been filed.
Launched in April 2026, Prentis focuses on "computer use models"—AI systems trained to navigate software interfaces and automate routine office workflows the way a human worker would. The company has signed contracts worth up to $50 million with customers including healthcare management service organizations and manufacturers.
Heavyweight Backers, Lean Team
Prentis was co-founded by CEO Ritankar Das, LinkedIn co-founder Reid Hoffman, and Zynga founder Mark Pincus. Das previously founded Titan Holdings, which builds and operates AI companies including Tala Health and Forta Health. Hoffman and Pincus are treating Prentis as a side project while focusing on other ventures.
Despite the high-profile founding team, Prentis operates with just over 25 employees. The team includes former researchers from OpenAI, Google DeepMind, Meta, Tencent, and Alibaba.
Entering a Crowded Field
Prentis enters a competitive market for computer-use agents. Major players including Anthropic, OpenAI, and Thinking Machines are also developing models capable of controlling computers and software applications. The lab's approach emphasizes a continuous loop of perceiving interfaces and improving from outcomes, rather than relying solely on web-scale training data.
The venture signals a broader shift in AI's value proposition: from chatbot or coding assistant to "action engine" capable of automating mundane, undocumented workflows across non-technical corporate roles. If successful, smaller and cheaper models that effectively control software could disrupt traditional business process automation.
Unverified Performance Claims
The company claims its Hive-32B model outperforms GPT-5.4 and Claude Opus 4.6 on WindowsAgentArena and ScreenSpot-v2 benchmarks. Prentis also states in its pitch deck that it achieves roughly 10 times lower cost per task than frontier APIs. Neither claim has been independently verified.
Investor materials project an estimated $75 million annualized run rate by Q3 2026. However, the company's pitch deck explicitly states these figures reflect estimated value based on 20% of customer savings—not recognized revenue—and are performance-dependent and subject to final execution.
What's at Stake
Prentis's bet is that AI agents capable of interacting with existing software interfaces can deliver productivity gains without requiring businesses to overhaul their technology stacks. The healthcare and manufacturing contracts suggest early traction, though specific customer identities remain undisclosed.
The success or failure of this approach could determine whether AI becomes a tool for automating the bulk of office work or remains confined to specialized technical tasks.