• 4 mins read
  • Published

Why Silicon Valley’s AI Agents Still Aren’t Mainstream

Paul Christiano Journalist FAYFO Media

by Paul Christiano

Why Silicon Valley’s AI Agents Still Aren’t Mainstream FAYFO Media © fayfo.com
Why Silicon Valley’s AI Agents Still Aren’t Mainstream © fayfo.com

Despite billions invested and rapid technical progress, AI agents remain largely unused by everyday consumers. Industry leaders say the tech world must shift focus to products people actually want, not just impressive demos.

Inside Silicon Valley, the belief that AI agents will shape the future is nearly universal. Engineers are building payment systems for agents, automating workflows, and even developing safeguards to prevent agents from breaching other organizations. Yet, for most people outside the tech bubble, AI agents remain unfamiliar and unused.

Josh Miller, CEO of The Browser Company, captured this disconnect in a widely shared X post, questioning why, despite the technology’s readiness, the general public “doesn’t give a fuck” about AI agents. In an interview, Miller emphasized that while the industry is excited about the potential for agents to transform daily life, there’s little evidence of real-world adoption. “I just have not heard a single person outside of the tech community talk about an agent that they use,” Miller said, urging the industry to focus on building products that genuinely appeal to consumers.

Recent data highlights the gap: OpenAI reports about 10 million weekly users for its Codex and ChatGPT Work agents, while Anthropic’s Claude Code and Cowork agents see similar numbers. In contrast, chatbots like ChatGPT and Gemini each attract around a billion monthly active users. For AI labs, this means agents are still a niche, not a revolution.

Most consumers currently use generative AI for simple tasks-searching for information or casual conversation. Silicon Valley, however, has invested heavily in models capable of much more, hoping agents will unlock new value. But so far, no “killer app” has emerged to drive mass adoption. This concern echoes broader industry anxieties about the lack of a breakthrough moment for agents, similar to what ChatGPT achieved for chatbots.

Product vs. Technology

Miller argues that the problem is not just a lack of interest, but a fundamental misunderstanding of what consumers want. “No one wants AI agents, because AI agents aren't a thing. It is an invented frame made up by our industry to collectively refer to something,” he said. Instead, he believes the focus should be on building products that deliver clear, tangible benefits-like The Browser Company’s AI-powered browser, Dia, which offers users a personalized morning briefing with a greeting, to-do list, and curated content. While powered by an AI agent, users don’t need to know or care about the underlying technology.

Many current agentic products, Miller contends, are designed to showcase technical capabilities-such as navigating websites or writing code-rather than solving real consumer problems. These features often feel more like demos than useful tools. He attributes this to groupthink within the AI sector, where builders are often guided by a sci-fi vision of the future, referencing films like “Her” as inspiration. “There is sort of a lack of diversity of opinions and convictions,” Miller observed.

Rethinking Adoption

The latest trend in AI agents is empowering users to automate personal tasks, from building pitch decks to organizing data. While tools like ChatGPT Work and Claude Cowork have proven useful for some, the audience for DIY productivity solutions remains limited. For agents to reach mainstream adoption, the industry needs to think bigger and design products that are joyful, useful, and approachable for a broad audience.

Miller hopes his message will resonate beyond major labs like OpenAI and Anthropic, encouraging founders and product teams to question assumptions and reimagine what agent-powered products could be. “Who gives a shit if it looks like an AI agent? No one uses AI agents,” he said, challenging the industry to focus on outcomes, not labels.

As the debate continues, some experts point to the rapid evolution of AI-powered security tools as an example of how practical applications can drive adoption. For instance, a recent report highlighted how AI-driven security solutions are already transforming business server protection, suggesting that consumer-facing breakthroughs may not be far behind if the right products emerge.

Josh Miller’s track record lends weight to his perspective. His previous startup, Branch, was acquired by Facebook in 2014, and his latest venture, Arc, was bought by Atlassian for $610 million. Miller also served as the White House’s first director of product under President Obama, giving him a unique vantage point on both technology and user needs.

Related articles