What Happened

Prentis, a new AI research lab focused on computer-use models, is in talks to raise $100 million at a $1 billion valuation, according to two sources familiar with the discussions who spoke with TechCrunch. The lab was launched in April 2026 and is co-founded by serial entrepreneur Ritankar Das alongside tech heavyweights Reid Hoffman (LinkedIn co-founder, Greylock partner) and Mark Pincus (Zynga founder, Reinvent Capital).

Prentis is training models to automate routine office workflows — navigating documents, handling insurance claims, processing customs duty refunds — by building AI agents that can directly control computers. According to investor materials obtained by TechCrunch, the startup has already signed contracts worth up to $50 million with customers including a healthcare management organization and several goods and clothing manufacturers. The company projects $75 million in annualized run rate by Q3 2026, though its own pitch deck notes these figures reflect estimated annualized value based on a contracted fee equal to 20% of savings realized and are "performance-dependent and subject to final execution."

The startup claims its Hive-32B model outperforms OpenAI's GPT-5.4 and Anthropic's Claude Opus 4.6 on two computer-use benchmarks: WindowsAgentArena (end-to-end task completion on real Windows applications) and ScreenSpot-v2 (locating the right on-screen control). Prentis says its edge is running a much smaller, cheaper model — roughly 10 times lower cost per task than frontier APIs. TechCrunch has not independently verified these benchmark results.

Prentis has hired more than 25 employees, including researchers previously at OpenAI, Google DeepMind, Meta, Tencent, and Alibaba. CEO Ritankar Das, 31, was UC Berkeley's youngest University Medalist in over a century, graduating at 18 before earning a master's at Oxford and dropping out of an AI PhD at Cambridge. He previously founded Titan, a Berkshire Hathaway-style holding company that builds and operates AI companies including Tala Health and Forta Health.

For Hoffman and Pincus, Prentis is a side project of sorts. Hoffman recently stepped down from Microsoft's board to go "founder mode" on Manas AI, an AI drug-discovery startup. He was an early OpenAI investor and co-founded Inflection AI before Microsoft absorbed most of that team in 2024. Pincus now runs Reinvent Capital with Hoffman as a senior adviser.

Why It Matters

Prentis is making a specific contrarian bet: that automating everyday office tasks will soon outpace coding as AI's biggest use case. This is a meaningful divergence from where much of the AI investment and developer attention has concentrated over the past two years. If the thesis is right, the total addressable market for computer-use agents could dwarf coding assistants — every back-office worker navigating multiple systems is a potential user.

The smaller-model approach is the more interesting technical claim. If a 32B model can genuinely match or beat frontier models on computer-use benchmarks at 10x lower cost, it challenges the prevailing assumption that computer-use agents require frontier-scale intelligence. This would have immediate implications for anyone building agent workflows on OpenAI or Anthropic APIs today — the unit economics could shift dramatically.

However, the competitive landscape is already intense. Anthropic, OpenAI, and Mira Murati's Thinking Machines Lab are all developing computer-use agents. Anthropic has been aggressively consolidating talent in the category, acquiring Seattle-based computer-use startup Vercept earlier this year and shutting down its product. Prentis is entering a market where well-capitalized incumbents are already acquiring competitors.

The $50M in signed contracts — if real and converting — is the strongest signal here. Enterprise buyers signing performance-based contracts for computer-use automation suggests the category is moving from demo videos to deployed workflows faster than many expected.

Who Is Affected

AI startups building computer-use or workflow automation agents now face a well-funded competitor with high-profile founders, enterprise traction, and a cost-disruption narrative. The performance-based pricing model (20% of savings realized) sets a precedent that buyers may come to expect from any vendor in this space.

Enterprise IT and operations leaders evaluating AI automation tools have another vendor to benchmark. The claims around insurance claims processing and customs duty refunds suggest Prentis is targeting specific vertical workflows rather than building a general-purpose agent.

Developers building on frontier APIs for computer-use tasks should monitor whether Prentis opens API access or remains enterprise-contract-only. If smaller specialized models deliver comparable results at a tenth of the cost, the economics of agent-based applications change fundamentally.

Strategic Implications

For AI startup founders

If you're building in computer-use or workflow automation, Prentis's $50M in signed contracts signals that enterprise buyers are ready to pay for this category now — but they'll expect performance-based pricing models. Price your offering as a share of savings or outcomes, not as a flat API fee. Differentiate on vertical depth rather than competing on raw model performance.

For developers/operators building with AI APIs

Prentis's claim of 10x lower cost per task with a 32B model versus frontier APIs is the number to watch. If they open access and the claim holds, your cost structure for agent workflows could drop significantly. Start benchmarking your own computer-use tasks against smaller open models now to understand where frontier APIs are overkill.

For non-technical business owners evaluating AI tools

Computer-use AI agents are reaching contract-signing maturity for document-heavy workflows. The performance-based pricing model means you can pilot with minimal upfront risk. But verify the vendor's actual deployment track record — Prentis's own pitch deck acknowledges its revenue projections are "performance-dependent and subject to final execution." Benchmark scores don't equal production reliability.

What to Watch Next

Monitor whether Prentis closes this funding round and at what valuation — the $1B figure is unconfirmed. Watch for any independent benchmark validation of Hive-32B's claims against GPT-5.4 and Claude Opus 4.6. Also track whether Anthropic or OpenAI respond with their own enterprise computer-use product launches, which could compress Prentis's window to establish market position.

Frequently Asked Questions

Q: What is Prentis AI and what does it do?

A: Prentis is an AI research lab co-founded by Reid Hoffman, Mark Pincus, and Ritankar Das that builds computer-use AI agents — models that can control computers to automate routine office workflows like insurance claims processing and customs duty refunds. The company claims its Hive-32B model outperforms frontier models from OpenAI and Anthropic on computer-use benchmarks at roughly 10x lower cost.

Q: How is Prentis different from Anthropic's computer-use agent?

A: Prentis is betting on a smaller, specialized 32B model approach rather than frontier-scale models, claiming this delivers comparable benchmark performance at a fraction of the cost. Prentis also targets specific enterprise verticals with performance-based pricing (charging ~20% of savings realized), whereas Anthropic's computer-use capabilities are offered through its general API. Anthropic has also been acquiring computer-use talent directly, including its purchase of Vercept earlier this year.