Town, a San Francisco startup building AI assistants for businesses, announced a US$108 million Series B on October 19, 2026, co-led by Forerunner Ventures and Menlo Ventures. According to TechCrunch, the round gives the company a unicorn valuation, meaning above US$1 billion, roughly four months after its public launch. The company has not published an exact valuation figure.
What was announced
TechCrunch reports that the Series B was co-led by Forerunner Ventures and Menlo Ventures, with participation from Andreessen Horowitz, Serena Ventures and First Round Capital. Total funding is reported as at least US$160 million, including the US$55 million Series A announced in June. The company says it reached "millions of dollars" in annual recurring revenue within weeks of launching, but gives no exact figure and no user count.
Town raised nearly twice as much in its second round, four months on
Round size in millions of US dollars. TechCrunch reports total funding of at least $160 million.
The two rounds shown total $163 million, consistent with the reported "at least $160 million" in total funding.
The two rounds shown add up to US$163 million, consistent with the reported "at least US$160 million". The Series B is nearly twice the size of the Series A, and it arrived about four months after the public launch.
How the product works
Each employee at a customer company gets an assistant called a "Townie". According to TechCrunch, Townies connect to email, calendars and other core work systems, can talk to one another and share approved information, and can now also work with customers and other outside parties. A new shared workflow updates every employee's Townie automatically when new business information arrives, so staff do not have to update their own assistant.
That last feature is the most analytically interesting. Most workplace AI tools treat the assistant as a personal tool: it knows what you tell it and what you connect. A shared update mechanism treats the company as the unit of context.
An analysis: three design choices that differ from a single chatbot
| Design choice | What it enables | What it requires |
|---|---|---|
| One assistant per employee | Personal context and preferences | A way to keep each assistant current |
| Assistant-to-assistant sharing of approved information | Handoffs without a human relaying context | Clear rules on what counts as approved |
| Shared business-information workflow | Consistency across the company | Strong permissions, so shared context does not leak across roles |
| Assistants working with outside parties | Customer-facing automation | Higher stakes for errors, since replies leave the building |
Each row trades convenience against governance. Connecting an assistant to email and calendars creates a rich context source, and it also creates an attack surface. We analysed how poisoned content can persist in agent memory in AI agent memory poisoning and prompt injection, and the same concerns apply wherever assistants exchange information with each other or with external parties.
What the valuation does and does not tell us
Reported facts allow only limited inference:
- Speed. Reaching a unicorn valuation roughly four months after launch is fast by historical standards, but the valuation itself has not been disclosed, so no multiple can be computed.
- Revenue. "Millions of dollars" in annual recurring revenue is a range spanning more than an order of magnitude, from US$1 million to US$9 million at the most literal reading. Any revenue multiple implied by a US$1 billion-plus valuation would therefore be very high, but the true figure cannot be derived from the public record.
- Investor overlap. Andreessen Horowitz and Forerunner appeared in the earlier round, so part of this round reflects existing investors increasing their position, which is a common pattern when a company is scaling quickly.
A valuation four months after launch prices a thesis about workplace agents, since the revenue figures disclosed so far are too coarse to price anything else.
Where this sits in the assistant market
Personal and workplace assistants are a crowded category. Large model providers sell assistants directly, platform companies embed them in productivity suites, and startups like Town compete on depth of integration and on how the assistants coordinate. Our earlier coverage of consumer-facing assistant startups, such as Underdog's on-device assistant, shows the same investor interest in assistants that sit close to a person's daily tools.
For teams evaluating this category, a few questions matter regardless of vendor:
- Which model does the assistant run on, and can it change? Locking an assistant to one model provider ties its quality and price to that provider.
- What does an assistant see by default? Read access to all email is a larger grant than access to a shared folder.
- How is shared context audited? If every assistant is updated automatically, someone needs to see what changed and when.
- What happens when an assistant acts externally? Approval steps for outbound messages are a basic control.
Tools built around model choice, such as Metir, approach the first question by letting users pick among providers per task. The wider point is that assistant quality and assistant governance are separate axes, and buyers need to assess both.
What to watch next
- A disclosed valuation or revenue figure, which would let observers evaluate the round properly.
- Customer-facing deployments. The ability to work with customers and outside parties is the highest-stakes feature, and its incident record will matter.
- Competitive responses from productivity suites that already hold employees' email and calendars.
Sources:
- AI assistant startup Town nabs unicorn valuation four months after launch (TechCrunch)
- Town raises $55M Series A for personalised AI assistants (Dealroom)
Image credits
- Hero and in-body photo: "San Francisco Downtown Aerial, August 2025" from Wikimedia Commons, released under CC0 (public domain dedication). Pulled 2026-10-21 and reviewed visually before use.