What Happens When Al Agents Can Spend Your Money?

The episode begins with Alex Rampell reminiscing about working with Jean-Denis Greze at Plaid and how Greze left to build a new company. Initially planning to develop SMB tax software, Greze discusses the challenges of market penetration and the realization that tax automation wouldn't scale quickly due to limited revenue opportunities and incumbent competition. After a year of building the tax product, Greze and his co-founder decided to pivot, motivated by their desire to create meaningful impact and the recognition that the tax market wasn't the right fit for rapid growth.

Greze shares how they shifted focus to building AI assistance around email, calendars, and routine workflows that dominate daily business activities. This pivot was driven by conversations with users who wanted AI help with meeting prep, email responses, and account management. The rapid advancement in AI models facilitated a quick product-market fit, and the team realized the potential to automate common knowledge worker tasks that are repetitive and mundane, thus enabling productivity and saving time.

The episode discusses how Town focused on an easy-to-use, self-serve product that connects with email and calendar, which are ubiquitous in business workflows. Greze emphasizes the importance of designing for the user, not just the buyer, and initially prioritized product excellence over enterprise-specific features. This approach aimed to create consumer-grade experiences within enterprise settings, making it easier for CEOs and CIOs to adopt Town's assistant by offering familiar, intuitive interactions and gradually expanding security and compliance as the product matured.

Greze explains Town's neutrality in AI model sourcing, using the best provider for each task rather than being tied to a single supplier like OpenAI or Anthropic. This strategy aims to de-risk the business from dependence on one vendor and maintain control over data ownership and model choice. Town also focuses on owning the data and context generated within its product, making switching costs low and fostering trust by ensuring customers own their contact and activity data, contrasting with other labs' more proprietary approaches.

The conversation shifts to the future impacts of AI agents acting proactively on behalf of users, managing business workflows, shopping, and even social activities. Greze asserts that widespread adoption will depend on making AI experiences as compelling and effortless as personal assistants, which involves deep integration with existing tools and data. He highlights the challenges of disseminating AI's benefits globally and across sectors, emphasizing how small innovative businesses and “tinkerers” are already unlocking value in unexpected ways. Greze calls for user-centered product design that minimizes friction and maximizes practical automation.

The episode explores how companies should think about the balance between horizontal products and vertical specialization in AI. Greze advocates starting broadly to gather diverse data and insights, then verticalizing based on user needs, especially as AI becomes better at general capabilities but more valuable when tailored. He discusses the strategic importance of neutrality, ownership of data, and avoiding lock-in with suppliers, contrasting Town’s approach with the very targeted bets of labs on coding, healthcare, or law. He emphasizes that AI's capacity to customize and learn from real-world use is the key to long-term success.

The final segment examines the dynamics of AI-powered agent traffic and how major platforms like Amazon and Meta could shift their strategies. Greze envisions a future where neutral, user-owned agents can operate without conflicts of interest—allowing, for example, Amazon’s effective use of agents to enhance shopping experiences without adversarial competition. He argues that proprietary, platform-controlled agents risk creating biases, and instead, open, neutral agents could unlock enormous value by enabling more seamless, personalized interactions—and ultimately, disrupting existing distribution, advertising, and search models.