Meta AI Leader Rohit Patel Transitions to Lead Microsoft Office and Google Workspace Initiatives

After an eight-year stint at Meta, Rohit Patel has officially left the company to pursue a new venture. He has founded Ndoc, a startup designed as a comprehensive AI-driven alternative to traditional office suites like Microsoft Office or Google Workspace.

Patel’s departure adds to a series of notable exits from Meta, which recently included Chief Global Affairs Officer Joel Kaplan, AI researcher Andrew Tulloch, and the Head of Product for AI-for-Work Emily Dalton Smith.

With Ndoc, Patel aims to go beyond simply adding an AI layer to existing workflows. The intention is to eliminate the disjointed nature of current tools such as Excel, Word, PowerPoint, Tableau, and others, by providing an all-in-one workspace accessible to all team members—even those who are not registered with Ndoc.

Ndoc’s design consolidates three primary components in a single interface. The Stage tab contains various presentation formats like slide decks, dashboards, and simulations, while the Narrative tab provides analyses, citations, and explanations of the data. The Model tab archives datasets and Python scripts used for generating documents.

Reflecting his expertise, Patel focused on reinforcement learning during his time at Meta. This machine learning method allows AI systems to learn from trial and error to make decisions. Patel insists that for AI-generated outputs to be utilized in business decision-making, they must be verifiable. Users can easily access the Python scripts that produce these documents.

“Even if you’re not familiar with Python, you can ask the AI, ‘What does this script do? Which line has my formula?’ You can confirm your formula easily,” Patel explained. “There’s transparency in the process.”

Users can upload datasets for analysis, and Ndoc will produce a narrative regarding the findings. Document-specific details like audience preferences, tone, and contextual memory are also incorporated. “It retains memory tailored to you, your document, your team, and your organization,” Patel noted.

Patel claims that Ndoc is built from a combination of open-source models. Because it was developed from the ground up, he argues that it operates at a much lower cost than competitors relying on advanced AI frameworks.

“Initially, I thought people would be thrilled about the cost-efficiency, but user excitement shifted to appreciating the comprehensive view of their projects,” he remarked.

Beyond just functioning as a tool, Patel envisions a new type of document that can be exported as a single “.ndoc.html” file compatible with any browser. He intends to make this format open-source so that developers can create applications that interact with it, avoiding the constraints faced by users of Microsoft or Google products.

Patel’s commitment to open-source solutions was a key reason behind his decision to leave Meta. He believed that reaching consensus on an open-source format in such a large corporation would involve numerous viewpoints and could slow progress.

“Humans operate at a human speed, whereas AI can accelerate things,” he emphasized.

Aiming to establish Ndoc as a pivotal platform for enterprise tasks is a significant endeavor for a founder with a small team and no funding, especially with so many other AI companies eyeing the same market.

Major companies like OpenAI, Anthropic, Meta, SpaceXAI, and Google have released enterprise AI tools that integrate with Office applications, making it easier for users to apply AI capabilities directly within those settings. Recently, OpenAI introduced Space, a collaborative workspace in ChatGPT that poses a direct challenge to established solutions.

While Patel recognizes the potential competition from Space, he critiques the broader approach taken by many AI labs.

“Companies are merely attaching AI functionalities to existing spreadsheets, and I find that disheartening,” he commented. “It’s akin to discovering electricity and using it just to add lights to oxen instead of utilizing a tractor.”

However, there’s a rationale behind these practices, as changing established workflows often proves challenging. Patel has designed connectors to allow users to bring data from their current tools into Ndoc easily.

“I set a high standard: I won’t release anything unless I genuinely enjoy using it,” he stated. “If I find myself reverting to my old workflows, I haven’t succeeded. I have a collection of failed tools that didn’t meet those standards.”

Patel is equally focused on validating customer interest before seeking investors. “I wanted to ensure there was proof of love for this product before accepting funding,” he said.

As of now, Ndoc is running an enterprise pilot and has opened an alpha version for a select group of customers on a waitlist. Users can sign up for free but will incur charges based on usage and token systems. Patel aims to launch a beta version soon, with a full rollout slated for next year.

For Patel, speed is more crucial than achieving perfection initially.

“Execution is relatively inexpensive… so proceed with your plans even if they aren’t perfect and refine later,” he remarked. “By the time you’ve shaped an ideal strategy, someone else might already have executed theirs.”