Beyond the Code: The Battle for the 'Last Mile' of Office Automation

For the last two years, the AI narrative has been dominated by the 'developer productivity' loop—AI writing code to build better AI. But as a practitioner, I've always suspected the real friction isn't in the syntax, but in the 'last mile': the grueling, manual labor of navigating clunky corporate software to get a single task done.

This is the specific bet being placed by Prentis, a new AI research lab co-founded by Ritankar Das, Reid Hoffman, and Mark Pincus. According to reporting from TechCrunch, Prentis is currently in discussions to raise $100 million at a $1 billion valuation. Rather than focusing on generative creativity, Prentis is training models to master 'computer use'—essentially teaching AI how to navigate the routine workflows that plague office workers.

**Opinion: The Execution Gap**

In my view, the industry is finally hitting the wall of 'chat-based' productivity. Asking a LLM to draft an email is easy; asking it to hunt down a specific customs duty refund exception across three different legacy systems and a PDF archive is where the actual business value lives. Prentis is targeting this exact friction. TechCrunch reports the startup is developing agents to handle high-friction tasks like insurance claims and customs duty refund exceptions, removing the need for humans to manually hunt for paperwork.

**The Efficiency Play**

To win this race, you don't need the biggest model; you need the most efficient one. Prentis claims its Hive-32B model outperforms frontier models like Anthropic’s Claude Opus 4.6 and OpenAI’s GPT-5.4 on two specific benchmarks: ScreenSpot-v2 (locating on-screen controls) and WindowsAgentArena (end-to-end task completion on Windows).

Crucially, Prentis argues in its pitch deck that its edge is cost. The company claims a cost per task roughly 10 times lower than frontier APIs, making it economically viable to deploy across the mundane, repetitive workflows that define the average workday.

**A Crowded Frontier**

Prentis isn't alone in this pursuit. TechCrunch notes that OpenAI, Anthropic, and Mira Murati’s Thinking Machines Lab are all developing computer-use agents. Anthropic has already moved aggressively into this space, acquiring the Seattle-based startup Vercept earlier this year to absorb its talent.

Despite the competition, Prentis is already seeing commercial traction. TechCrunch reports the startup has signed contracts worth up to $50 million with a manufacturer, clothing manufacturers, and a healthcare management service organization. Investor materials obtained by TechCrunch predict an estimated $75 million annualized run rate by Q3 of this year, though the company notes these figures are based on a contracted fee of 20% of realized savings and are performance-dependent.

**The Architect**

The venture is led by CEO Ritankar Das, who founded the AI holding company Titan. Das’s track record includes launching Tala Health (which raised a $100 million seed round) and Forta Health (which raised $55 million led by Insight Partners in 2024), as well as Dascena, which was acquired by CirrusDx in 2022.

If Prentis succeeds, it will prove that AI's biggest impact isn't in replacing the coder, but in liberating the office worker from the digital drudgery of the 'last mile.'