The current AI gold rush is largely defined by the Copilot. From coding assistants to slide-deck generators, the prevailing narrative in Silicon Valley is one of productivity gains and marginal efficiency. Developers and enterprises are racing to integrate LLMs into existing workflows to shave off a few hours of manual labor per week. This is the era of the additive tool, where AI is treated as a sophisticated layer of polish applied to old, clunky processes. However, a quiet but profound shift is occurring among the most successful venture capitalists, who argue that the real value of AI lies not in making old processes faster, but in making them irrelevant.

The USV Philosophy of Structural Replacement

Fred Wilson and Union Square Ventures (USV) have long operated under a rigorous investment lens known as Don't Automate, Obliterate. This principle, which draws inspiration from a 1990 Harvard Business Review article by Michael Hammer, posits that the greatest economic opportunities do not come from automating existing tasks, but from obliterating the entire structure that necessitates those tasks. In the eyes of USV, a company that uses software to make a legacy process 20 percent more efficient is a feature, while a company that redesigns the market so the process is no longer needed is a platform.

This distinction is not merely semantic; it is a matter of power dynamics. Automation preserves the existing hierarchy while increasing the speed of the conveyor belt. Obliteration changes who owns the conveyor belt and who decides where it leads. USV's portfolio history provides a blueprint for this approach. Duolingo did not attempt to build a more efficient administrative system for language schools; it obliterated the need for the traditional classroom by providing a direct-to-consumer learning model. Coinbase did not simply digitize the ledger of traditional banks; it replaced the financial intermediary with programmable currency. Twitter did not create a faster way to submit press releases to publishers; it obliterated the publisher's role as the gatekeeper of information, allowing for direct, real-time public broadcasting.

In each of these cases, the goal was not to reduce the cost of a transaction within a market, but to fundamentally alter the distribution of power between the provider and the consumer. The success of these companies stemmed from their ability to identify a structural bottleneck—the school, the bank, the editor—and design a system where that bottleneck ceased to exist.

From Productivity Tools to Expertise Packaging

As the industry pivots toward AI agents, the Don't Automate, Obliterate framework becomes the primary differentiator between a failing SaaS tool and a category-defining company. The core tension in the AI market today is the difference between providing a tool for a professional and packaging the professional's expertise into a deployable agent. When a company provides a tool to a lawyer or a doctor, they are automating. When they package the intelligence of a lawyer or a doctor into an agent that the user can deploy directly, they are obliterating the intermediary.

This shift represents a transition from a supplier-centric market to a demand-centric market. For decades, professional expertise has been guarded by institutions that controlled access through certifications, referral networks, and appointment calendars. The value of the intermediary was not just their knowledge, but their role as the gatekeeper of that knowledge. AI agents break this monopoly by commoditizing the delivery of expertise. The user no longer needs to rent intelligence from a firm; they can instantiate it via an API.

Consider the case of Doctronic. Most health-tech startups focus on automation: they build better patient portals, smoother scheduling apps, or digital intake forms. These tools make the waiting room more efficient, but the waiting room still exists. Doctronic takes a different path by providing AI physicians directly to smartphone users, bypassing the referral and waiting list structure entirely. In Utah, where the legal framework has allowed AI physicians to write prescriptions, this is not an improvement of the healthcare delivery process—it is the obliteration of the traditional primary care gatekeeper. The expertise is moved from the clinic into the pocket of the patient.

Similarly, Isembard is redefining high-precision manufacturing. The traditional approach to digitizing a factory involves selling software to existing CNC operators to help them calculate quotes or manage workflows more efficiently. This is automation. Isembard instead packages the operational expertise of precision manufacturing into an AI system and uses it to rebuild the factory from the ground up. By transferring the expertise from the human operator to the software, they remove the dependency on legacy facilities or specialized overseas labor. They are not adding a tool to the factory; they are replacing the factory's operational logic.

Even the realm of entrepreneurship is seeing this shift through companies like Cofounder. While many AI startups offer a suite of tools for marketing, sales, and product development—essentially a digital toolbox for a founder—Cofounder positions itself as the entity that actually launches and operates the business on behalf of the client. It does not sell a Copilot to help a founder work faster; it acts as the operator, obliterating the need for the founder to hire a full team of expensive specialists in the early stages. The AI is not the assistant; it is the agent of execution.

For any developer or founder building in the AI space, the litmus test for their business model is simple. If the service improves the booking, referral, or approval process of a middleman, it is automation. If the service allows the user to bypass that process and access the intelligence directly, it is obliteration. The former creates a tool that can be easily replaced by the next update to a large language model; the latter creates a new market structure that captures the value previously held by the gatekeeper.

The era of the professional intermediary is ending, replaced by a world where expertise is a deployable utility rather than a guarded secret.