Former Meta AI Director Launches Ndoc to Challenge Office and Workspace

After eight years at Meta, AI executive Rohit Patel is leaving the company to build a startup that aims to change how businesses create documents, analyze data, and collaborate. His new venture, Ndoc, is designed as an AI-native alternative to Microsoft Office and Google Workspace.

TechCrunch exclusively reported Patel’s departure from Meta, where he previously served as a director at Meta Superintelligence Labs. He is among several high-profile executives and researchers to have left the company, following the exits of Chief Global Affairs Officer Joel Kaplan, AI researcher Andrew Tulloch, and Head of Product for AI-for-Work Emily Dalton Smith.

Rather than adding AI features to existing workplace software, Patel wants Ndoc to bring the entire process into one connected environment.

Ndoc Aims to Bring Workplace Tools Into One Platform

Businesses often switch between spreadsheets, presentations, documents, analytics platforms, coding environments, and AI chatbots to complete a single project. Patel believes this fragmented workflow creates unnecessary complexity.

Ndoc aims to combine tasks commonly performed using Excel, Word, PowerPoint, Tableau, Replit, and ChatGPT into a unified workspace. The platform is also designed to make its work accessible to colleagues who do not have Ndoc accounts.

The product organizes its features into three main sections:

• Stage: Presents key findings through slide decks, dashboards, simulations, and small applications.
• Narrative: Explains the results with analysis, numbered citations, generated figures, and supporting context.
• Model: Stores the underlying datasets and Python scripts used to produce the work.

This structure is intended to connect the results of an analysis with the reasoning and data behind them, allowing users to examine the complete process rather than simply receiving an AI-generated answer.

Transparency Is Central to Ndoc’s AI Approach

Patel’s experience at Meta helped shape Ndoc’s emphasis on transparency and reliability. His work at the company focused on reinforcement learning, a machine-learning approach in which AI systems learn to make decisions through trial and error, as well as evaluating the trustworthiness of AI-generated output.

He believes businesses need ways to verify AI-generated spreadsheets, presentations, and other materials before using them to make important decisions.

For example, Ndoc allows users to inspect the Python scripts behind generated documents. Even people without programming experience can ask the platform’s AI chat to explain what a script does or identify the line containing a particular formula.

“There’s nothing hidden behind the curtain here,” Patel said.

The idea is to make AI-generated work easier to understand and verify instead of asking users to trust the final result without examining how it was produced.

Users can upload datasets, request an analysis, and have Ndoc develop a written explanation of the findings. They can also provide additional instructions about the intended audience, writing tone, and other document-specific requirements.

According to Patel, the platform maintains context across individual users, documents, teams, and entire companies. This memory is intended to help the AI produce more relevant results as projects develop.

Startup Uses Open-Source Models to Reduce Costs

Ndoc is built and fine-tuned using a combination of open-source AI models. Patel says developing the platform from scratch has also helped keep operating costs considerably lower than those of alternatives that rely on frontier AI models.

However, early users have shown him that affordability is not necessarily the product’s biggest selling point.

“I thought that’s what people would be excited about, but people don’t care about that,” Patel said, describing feedback from early trial users. He added that customers are more excited about having their entire workflow available in one place.

Patel’s longer-term vision goes beyond building another productivity application. He wants to change the way digital documents themselves work.

Ndoc documents can be exported as a single .ndoc.html file that opens in a web browser. Patel plans to open-source this format so developers can build compatible applications without being dependent on Ndoc alone.

He sees this as a potential alternative to the dependence many businesses have developed on Microsoft and Google software.

Why Patel Chose to Leave Meta

Patel says the desire to create an open document format was one reason he decided to leave Meta and pursue the project independently.

At a large company, establishing an open-source standard could involve extensive discussions and competing opinions about how the format should work. As a founder, he believes he can make decisions and build the product much faster.

“Humans can only move at human pace, whereas…I can move at the pace of what AI can do,” he said.

That speed could become an advantage, but Ndoc faces a significant challenge. Patel currently has a team of three and has not raised funding, yet he wants to compete in the enterprise productivity market.

Several major AI companies are already pursuing the same customers.

OpenAI, Anthropic, Meta, SpaceXAI, and Google have introduced enterprise AI tools that connect their models with workplace applications such as Google Workspace and Microsoft 365. These integrations allow users to bring AI capabilities into software they already use.

Last month, OpenAI also launched Space, a shared workspace inside ChatGPT that more directly competes with established workplace platforms.

Patel recognizes that products such as Space could pose a threat. However, he believes the broader industry is too focused on adding AI features to existing software instead of redesigning how people work.

“Companies are basically attaching AI to spreadsheets and it fills your cells for you,” he said. He compared that approach to putting lights on bulls used for farming when the better solution would be to introduce a tractor.

His argument is that AI should transform the workflow itself rather than simply automate individual tasks within familiar applications.

Connecting Existing Tools Could Help Ndoc Gain Users

Replacing established workplace software is difficult because businesses have already invested in their existing systems, processes, and employee habits.

Patel acknowledges this challenge. Ndoc includes connectors that allow customers to import existing data from tools they already use, providing a potential starting point for organizations that do not want to abandon their current workflows immediately.

He also follows a personal rule when developing the product: he does not want to release something he would not choose to use himself.

“My bar was that I’m not going to release anything I don’t love using,” Patel said, explaining that he already has highly optimized AI workflows for his own work. If he continues returning to those workflows instead of using Ndoc, he considers that a sign the product needs further improvement.

The same principle applies to fundraising. Patel says he wanted evidence that customers genuinely valued the product before accepting outside investment.

Ndoc Begins Enterprise Testing Ahead of Wider Launch

As of Tuesday, Ndoc is running an enterprise pilot and opening an alpha version to a limited number of customers on its waitlist.

Signing up is free, but customers will pay for their usage and tokens. Patel hopes to introduce a beta version in the coming weeks, followed by a full launch next year.

For now, the startup is betting that customers will embrace a different approach to workplace productivity, one that brings analysis, explanations, and finished deliverables together rather than spreading them across multiple applications.

The challenge will be convincing businesses to adopt a new platform in a market dominated by established software companies and increasingly powerful AI tools.

Patel, however, believes rapid execution matters more than perfect planning at the beginning.

“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.”

Also read : inDrive Plans 2027 Robotaxi Tests in Kazakhstan With Pony.ai