Meta AI Director Rohit Patel Exits, Launches Ndoc to Replace Office

Meta Superintelligence Labs director Rohit Patel is leaving the company after eight years to launch Ndoc, a startup he pitches as an AI-native replacement for the Microsoft Office suite and Google Workspace, TechCrunch exclusively reported.

Patel joins a short list of high-profile Meta exits. Chief Global Affairs Officer Joel Kaplan, AI researcher Andrew Tulloch, and Head of Product for AI-for-Work Emily Dalton Smith have all left the company in recent months.

Patel isn't layering another AI copilot onto Word or Excel. He wants to replace the fragmented stack — Excel, Word, PowerPoint, Tableau, Replit, and what he calls "ChatGPT copy-paste" — with one unified workspace that colleagues can access even if they never sign up for Ndoc.

How does Ndoc actually work?

The product has three tabs in a single layout. The Stage tab holds presentations of key findings: slide decks, dashboards, simulations, or mini apps. The Narrative tab carries the analyses — numbered citations, generated figures, and the "why" behind the numbers. The Model tab stores the underlying datasets and the Python scripts that produced them.

That architecture traces back to Patel's Meta work. There, he focused on reinforcement learning — machine learning where AI agents learn decisions through trial and error to maximize rewards — and on evaluating whether AI output can be trusted. His position: if AI builds the spreadsheets and decks that companies use to make decisions, someone has to be able to check the work. Ndoc users can pull up the exact Python scripts behind any document.

"Even if you don't know any Python, you can ask [the AI chat] to explain, 'Hey, what does this script do? Which line has my formula?' And you can verify that formula," Patel said. "There's nothing hidden behind the curtain here."

In practice, users upload a dataset and ask Ndoc to run an analysis and write a narrative on the findings. They can add document-specific context — audience, tone, and memory. "It maintains a memory across your person, that particular document, your team, and your company," Patel said.

The takeaway: Ndoc splits every piece of work into what you show, why it's true, and how it was computed.

Why run on open-source models?

Patel built and fine-tuned Ndoc on a mixture of open-source models. Because the platform was developed from scratch, he says it costs considerably less to run than alternatives built on frontier AI models.

He expected that cost edge to be the selling point. His early trial users disagreed.

"I thought that's what people would be excited about, but people don't care about that," he said. "What people are excited about is that they love seeing all their end-to-end in one place."

The bigger ambition is reinventing the file itself. Documents export as a single ".ndoc.html" file that opens in any browser, and Patel plans to open source the format so other developers can build apps on it — without the lock-in that has kept businesses tied to Microsoft and Google for years.

That open-source drive is partly why he left Meta. Getting a giant company to agree on an open format would require the input of "hundreds of people and opinions on how the format should look." Patel wanted speed. "Humans can only move at human pace, whereas…I can move at the pace of what AI can do."

Can a team of three take on OpenAI and Google?

The bet is enormous. Patel has a team of three and no funding, while the largest AI labs are racing to own the same territory.

OpenAI, Anthropic, Meta, SpaceXAI, and Google have all shipped enterprise AI tools that connect to Google Workspace, Microsoft 365, and other workplace apps, putting their models directly inside the software people already use. Last month, OpenAI launched Space, a shared workspace inside ChatGPT and a more direct shot at the incumbents.

Patel concedes Space could be a threat. But he argues the labs have the wrong idea.

"Companies are basically attaching AI to spreadsheets and it fills your cells for you," he said. "It just sort of kills my soul. This is like if you just figured out electricity, and then you decided the best use case is to put lights on my bulls that till my soil. No. Get a tractor."

The labs have their logic, though: changing habits and ripping out whole workflows is brutally hard. Patel gets that, which is why Ndoc ships connectors so customers can import existing data from the tools they already use.

He held the product to a personal standard before launch. "My bar was that I'm not going to release anything I don't love using," he said. "I have very optimized AI workflows for everything I do. If I start using this tool, and I keep going back to my workflows, I haven't built a good enough tool…I have a graveyard of tools that did not meet that bar."

The same bar applies to fundraising. "I didn't want to take money until I had some proof that people love the product," he said.

What ships now, and when?

As of Tuesday, Ndoc is running an enterprise pilot and opening an alpha version to a limited number of wait-list customers. Signing up is free; customers will pay for usage and tokens. Patel hopes to release a beta in the coming weeks, with a full launch planned for next year.

He's explicit that speed beats perfection. "Execution is extremely cheap…so go execute on a suboptimal plan and worry about it later," he said. "By the time you make the best plan, the execution will be done by somebody else."

The race is now whether a three-person team with no funding can establish an open document format before OpenAI, Google, and Microsoft lock enterprises into their own AI workspaces.