Former Meta AI Director Launches Ndoc to Challenge Office Suites with Agent‑Driven Automation

Rohit Patel, the veteran behind Meta’s AI evaluation platform and reinforcement‑learning pipelines, is exiting the company after eight years to launch Ndoc, a venture that promises to “take on Microsoft Office Suite and Google Workspace” with a new breed of AI agents. The move, reported by TechCrunch, underscores a growing trend: seasoned AI talent is spinning out of big tech to build lean, agent‑centric products that directly address enterprise productivity bottlenecks.

Patel’s track record at Meta includes building the internal benchmark suite that measured large‑scale language models against human‑like reasoning tasks. That experience gives him a rare insight into the gap between headline‑grabbing model capabilities and the gritty reality of integrating those models into day‑to‑day workflows. Ndoc’s pitch deck, obtained by our sources, frames the problem as “context‑drift” in traditional office apps: users must constantly re‑enter data, switch between documents, and manually synthesize insights. Patel’s solution is an AI‑agent layer that lives inside the suite, autonomously surfacing relevant data, drafting documents, and even negotiating calendar conflicts.

From a capital perspective, Patel is courting a hybrid investor set—venture firms focused on AI infrastructure and strategic corporate backers seeking a foothold in the next wave of productivity automation. Early indications suggest a seed round in the $15‑$20 million range, with a post‑money valuation near $80 million. If the round follows a typical 20‑25% founder equity dilution, Patel will retain roughly 70‑75% of the post‑money cap table, a healthy stake for a founder‑first play. The capital efficiency of the model—leveraging Meta‑trained embeddings and open‑source RL‑HF tools—means burn rates can stay modest while the team builds a minimum viable agent that integrates via APIs with Office 365 and Google Workspace.

Strategically, Ndoc’s emergence could accelerate the “agent‑first” paradigm that analysts have been watching since the release of OpenAI’s GPT‑4‑Turbo. By focusing on the enterprise segment, Patel sidesteps the crowded consumer‑AI space and taps a market with multi‑year contracts and high switching costs. If Ndoc can demonstrate measurable productivity gains—say, a 15‑20% reduction in time‑to‑insight for knowledge workers—it will force incumbents to either acquire or partner with similar startups, potentially reshaping the competitive landscape of office software.

The broader AI ecosystem should read Patel’s move as a signal that deep‑tech talent is increasingly confident in building capital‑efficient, domain‑specific agents that solve concrete business problems, rather than chasing vanity metrics. For founders, the lesson is clear: align agent capabilities with real workflow friction points, and investors will follow.