AI Email Assistant Town Raises $108M Series B at $1B Valuation

U.S. venture capital firm Menlo Ventures announced on October 19, 2026, that it has co-led AI startup Town's $108 million Series B funding round alongside Forerunner, with participation from a16z and First Round Capital. The new valuation places Town at $1 billion.

Town's core product, an AI assistant named "Townie," is designed around a "work-first" philosophy rather than traditional conversational interaction. According to Menlo Ventures, Townie reads users' calendars and inboxes to proactively propose tasks, draft replies, and track follow-ups without requiring any user prompts. This approach aims to address the efficiency bottleneck where knowledge workers spend more than 13 hours a week on email, chat, and text messages.

Menlo Ventures stated that its entire team began using Town fully in June 2026. After five months of internal testing, Menlo reported that Townie demonstrated significant value in catching forgotten commitments, integrating context for replies, and coordinating meeting times. Additionally, Townie features cross-user collaboration capabilities, allowing one user's Townie to communicate directly with another's to schedule meetings without human intervention.

Regarding technical architecture and security, Town was founded by Jean-Denis Greze, former CTO of Plaid. The company is SOC 2-certified and commits to not training models on customer data. By default, Townie generates drafts rather than sending emails directly, a design choice intended to build user trust. Currently, Town integrates with over 60 tools, spanning CRM (e.g., Salesforce), project management (e.g., Jira), coding (e.g., GitHub), and finance software (e.g., QuickBooks).

This funding round reflects capital market recognition of the "proactive AI agent" niche. Unlike Copilot-style tools that rely on user instructions, Town attempts to redefine the boundaries of human-AI collaboration by deeply embedding itself into workflows. Menlo Ventures partner Shawn Carolan noted that while early similar attempts (such as Handle) failed due to technological limitations, current large model capabilities have made this type of "frictionless" automation feasible.