The Agentic Gap: Why AI Demos Aren't Converting to Commerce
AI-generated image · US National Wire
Silicon Valley is obsessed with the technical capabilities of AI agents, but as Josh Miller notes, the general public remains indifferent because the industry is building for the 'wow' factor rather than consumer utility.
In the corridors of Silicon Valley, AI agents are viewed as the inevitable future of productivity. Developers are already engineering payment systems specifically for these agents and utilizing them to automate professional workflows. However, as Wired first reported, there is a stark disconnect between this industry optimism and actual consumer behavior.
According to Wired, the adoption numbers tell a sobering story. While chatbots like Gemini and ChatGPT average around one billion monthly active users, the agentic offerings are a fraction of that. OpenAI reported that its ChatGPT Work and Codex agents have roughly 10 million weekly users, and sources close to Anthropic indicate that its Cowork and Claude Code agents are seeing similar levels of adoption. In the context of mass-market retail and e-commerce, these figures are essentially a rounding error.
***Opinion:*** *From a commerce operator's perspective, this gap exists because developers are treating agents as the product rather than the engine. We are seeing a recurring pattern where tech labs prioritize the 'demo'—the ability to navigate a website or write complex code—over solving the granular friction points of a consumer's daily journey. Until an agent solves a tangible pain point in the checkout or discovery process, it remains a novelty, not a utility.*
Josh Miller, CEO of The Browser Company, has become a vocal critic of this approach. In a viral X post, Miller noted that while the technology is theoretically ready to transform lives, the general public simply does not care. Miller argues that "AI agents" are not a product, but rather an invented frame used by the industry to describe a set of capabilities. He suggests that the goal should be to create products that make users feel focused and calm, regardless of whether a "harness" calling up tools is powering the experience.
Miller points to the success of Dia, an AI-powered browser from The Browser Company, as a blueprint. Its most popular feature is a personalized morning briefing that aggregates a user's calendar, email, and curated content. While Miller acknowledges this feature is technically powered by an AI agent, he emphasizes that the user does not need to know that to find value in the product.
This misalignment is driven, in part, by what Miller describes as industry groupthink. He told Wired that during meetings with leaders at top AI labs, nearly every executive referenced the 2013 film *Her* as the guiding vision for their products. This sci-fi aspiration often overrides the practical needs of the average consumer.
Currently, the industry is leaning into tools that allow users to build personal software for tasks like organizing datasets or creating pitch decks. While these are useful for a niche group of productivity enthusiasts, they lack the broad appeal necessary for mainstream adoption. For AI agents to move beyond the demo phase, the industry must stop asking what the models *can* do and start asking what the consumer actually *wants*.