Town, a San Francisco startup building a personalized AI assistant that works across a user's email, calendar, messages, and work apps, has raised a $55 million Series A led by Andreessen Horowitz, with Forerunner Ventures, First Round, Alt Capital, and Conviction also participating. The deal, announced June 3, matters to any knowledge worker who has watched the AI boom without seeing their own workday change: Town's bet is that the assistant should learn you and act on your behalf, rather than wait for you to learn it.
Andreessen Horowitz Leads $55 Million Round Into Town
The round, reported first by Fortune, was led by Andreessen Horowitz, whose general partner Alex Rampell joined Town's board. In the company's funding announcement, Rampell argued that for consumer software "the moat isn't the model" but the context a product accumulates about a user, an advantage he said widens the longer someone uses it. Forerunner founding partner Kirsten Green framed the next wave of AI adoption as technology that quietly learns a person and works alongside them, rather than one more app to open. Town is chasing an AI-assistant market that Fortune, citing industry estimates, valued at roughly $16 billion in 2024 and projected to reach $74 billion by 2033, layered atop a productivity-software business worth about $110 billion today.
Town Pivoted From Small-Business Tax Software to Personal Assistant
Town's reinvention is part of the story. The company was founded in late 2024 by Jean-Denis Greze, former chief technology officer of fintech company Plaid, and Tony Vincent, a former director of applied AI at Google who previously led product design at Dropbox. Its first product was narrow: an AI-driven tax-and-compliance service for small businesses, built to give them the kind of tax advantages large corporations enjoy, and it launched with an $18 million seed round led by First Round Capital in early 2025. By mid-2026 the framing had widened into a general personal assistant, with tax becoming one capability among many. Greze describes the target user as a "prosumer," with roughly 70 to 80 percent of usage tied to work; one cohort is a community of Australian plumbers, one of whom told the company he fields 300 emails a day, and Town sorts all of it.
Connecting Inbox and Calendar Is How Town Learns Users
The product's design begins with connection rather than commands. A user cannot sign up without linking their email and calendar, and from there Town builds a working picture of who they are. In the first few minutes it surfaces a short biography assembled from what it can already see, then offers a handful of actions it can take immediately, such as drafting a research brief on an unfamiliar email sender or translating a document into a preferred language. Over time it learns preferences, notices patterns, and begins asking whether it should simply handle recurring tasks on its own. The company says Town is also being used for recruiting pipelines, logistical coordination, summarizing information, and drafting follow-ups.
Persistent Context, Not Base Models, Is Town's Intended Moat
Strip away the mascots and the wager is technical. Foundation models from OpenAI, Anthropic, and Google are converging in capability and falling in price, which means the model itself is becoming a commodity input rather than a durable advantage. Town sits on top of those models and on Google and Microsoft infrastructure, and treats the defensible layer as the persistent, per-user context it accumulates: identity, preferences, working rhythm, and relationships, stored and reused so the assistant improves with every interaction. Greze argues that even fairly engaged people open ChatGPT or Claude only about three times a day and lack intuition for what the tools can do, precisely because those systems are not wired into everything they use. The engineering tradeoff is the heart of the bet, and its catch: the same accumulated context that makes the assistant useful also concentrates a single user's most sensitive information in one place.
Gemini and Copilot Hold Deeper Native Data Access
That tradeoff defines the competition. Town faces general-purpose assistants from OpenAI and Anthropic with stronger base models but shallower workflow integration; vertical tools aimed at email, sales, or recruiting; and, most pointedly, the assistants built into the platforms it depends on. Google's Gemini and Microsoft's Copilot have far deeper native access to the same email and calendar data, and they ship inside the suites users already pay for. Greze does not dismiss the platform-dependency risk but cites precedent, noting that Superhuman has long competed with Gmail without being shut out and that Notion built a large business on top of Google Docs. Whether Town can sustain an open enough ecosystem on infrastructure it does not own is, by the founders' own admission, the long-term question. Its early retention data is the counterargument: the company says it is approaching 10,000 users, with 99 percent two-month retention among those who have built at least one custom automation.
Can AI Assistants Read Your Email Safely?
The deepest challenge is trust, and it is the one every agentic product shares. Software engineer Martin Fowler, writing about agentic email, called the prospect appealing but deeply scary: an inbox is a nerve center of sensitive data, and giving an agent direct access triggers what he calls the "lethal trifecta" of untrusted content, sensitive information, and the ability to communicate externally, the conditions under which a prompt-injection attack can quietly exfiltrate data. Regulators are circling the same problem. In a January 2026 report, the UK Information Commissioner's Office warned that personal-assistant agents concentrating communications and calendars raise surveillance and breach risks, and stressed that responsibility for protecting that data cannot simply be pushed onto end users. For Town, reliability, permission scoping, and human oversight are not features to add later; they are the wall it must build to turn a promising tool into a durable work entry point.
Frequently Asked Questions
What does Town's AI assistant do?
Town connects to a user's email and calendar and learns their habits, then triages the inbox, schedules meetings, drafts follow-ups, writes research briefs on unknown contacts, translates documents, and handles administrative work. The goal is for the assistant to act proactively rather than wait for prompts.
Who founded Town?
Town was cofounded in late 2024 by Jean-Denis Greze, former chief technology officer of Plaid, and Tony Vincent, a former applied-AI director at Google who also led product design at Dropbox. Greze is chief executive.
How is Town different from ChatGPT or Gemini?
General chatbots wait for instructions and are not wired into a user's tools. Town's pitch is the opposite: it plugs into your apps, remembers who you are across sessions, and bets that accumulated personal context, not a better model, is the durable advantage.
Is it safe to give an AI assistant access to your email?
It carries real risk. Security experts note that an agent with inbox access can be manipulated through hidden instructions to leak data, and privacy regulators warn that such tools concentrate sensitive information. Users should weigh an assistant's permission controls and human-oversight safeguards before granting access.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
