Key Takeaways

  • Town secured a $108 million Series B led by Forerunner Ventures and Menlo Ventures at a $1 billion valuation to build enterprise AI agents.
  • Jean Denis argues that workers will stop navigating standalone applications, relying instead on personal assistants that optimize for their goals.
  • Most workplace tools trap AI models in private 1:1 prompts, whereas Town embeds agents into multiplayer threads where teammates collaborate.
  • Enterprise agents cannot strip decision-making power from humans early; they must earn trust through thorough research rather than rushed suggestions.

Software Is Moving Past the App Store

For decades, business software forced employees to adapt to specialized interfaces. You opened a browser tab for payroll, another for project tracking, and a third for customer tickets. Jean Denis believes that pattern is reaching its end. “We believe that the way most humans will be interacting with digital information is not going to be through apps in the future,” Denis said. “It's going to be through assistants that are very much theirs... and that are optimizing for their success in life.”

Instead of logging into dozens of SaaS portals, employees will interact with dedicated agents that handle retrieval, synthesis, and execution across channels. To back that premise, Forerunner Ventures and Menlo Ventures led a $108 million Series B round, valuing Town at $1 billion. Denis designed the platform to operate across text, email, and internal team threads, keeping enterprise data isolated from consumer channels to satisfy corporate security standards.

Why AI Must Live in Multiplayer Channels

Most generative AI products isolate the operator in a private conversation. You open an empty input box, type a prompt, evaluate the output alone, and copy the text into another tool to show your team. Denis views that design as an obstacle to organizational adoption.

“We've really focused on the multiplayer,” Denis explained. In real companies, work rarely happens in isolation. It takes place in collaborative threads, handoffs, and team review cycles. When an agent participates directly in a multiplayer channel, every colleague can inspect the research, assess the context, and verify the output. Shared visibility stops different departments from duplicating identical queries and helps teams establish common operating standards for automated work.

Trust Must Come Before Automation

Many enterprise automation platforms fail because they attempt full autonomy before proving competence. They trigger noisy notifications, draft unvetted messages, and interrupt users with unhelpful summaries. Denis warns that aggressive autonomy alienates workers.

“You have to earn the trust to become the one glass pane for people's universe,” Denis noted. “That's the goal. But you can't take away the control of the human without earning the trust first.”

To build that confidence, software must mirror the discipline of a competent coworker. “That feels more like what a good business partner would do,” Denis argued. “They do the research first. They're not just slamming you with a thing that they spent two minutes thinking about.” If an agent acts before it understands context, users shut it off. Denis believes founders building workplace agents must leave humans in control during early stages, proving accuracy on background research tasks before asking for permission to automate actions.

What to Do With This

Audit your team's internal bots and automated workflows tomorrow morning. Identify every tool that pushes unsolicited recommendations or autonomous actions into your channels. If a system triggers actions without showing its underlying research, strip its write access and restrict it to read-only research until its accuracy is proven across thirty consecutive days.