Institutional AI represents a fundamentally different class of technology from individual productivity tools, according to a March 12, 2026 essay by George Sivulka, a partner at venture capital firm Andreessen Horowitz (a16z). Sivulka argues that while AI has made individual workers roughly ten times more productive, organizations have not seen equivalent gains in value or revenue.
Sivulka draws a direct parallel to the electrification of American textile mills in the 1890s. Factories that installed electric motors in place of steam engines saw almost no increase in output for thirty years. Meaningful productivity gains only arrived in the 1920s, when manufacturers completely redesigned factory floors around the new technology. The lesson, Sivulka states, is that technology and institutional design must evolve together.
The Seven Pillars of Institutional Intelligence
Sivulka identifies seven factors that separate Institutional AI from individual tools. He argues the entire B2B artificial intelligence sector will be built on these distinctions over the next decade.
Coordination and Signal
The first pillar is coordination. Sivulka states that deploying AI agents without a coordination layer produces chaos, as each employee develops separate workflows whose outputs do not connect. He predicts an entire “Agentic Management” industry will emerge to define agent roles, agent-to-agent communication, and methods for measuring agentic value.
The second pillar is signal. As AI tools allow individuals to generate content at scale, Sivulka argues the volume of low-quality output has grown so large that some organizations have banned AI-generated work entirely. He states that Institutional AI must be deterministic and auditable, filtering noise to surface actionable information rather than simply accelerating content creation.
Bias, Edge, and Outcomes
The third pillar addresses bias. Sivulka states that current consumer AI models are over-aligned to agree with users, which he describes as organizationally toxic. He argues that Institutional AI must challenge assumptions, surface risks, and enforce standards, functioning as an auditor rather than an assistant. He cites AI board members, AI auditors, and AI compliance tools as consequential future applications.
The fourth pillar is edge. Sivulka contends that purpose-built, domain-specific solutions will always maintain a capability advantage over general-purpose foundation models for specific tasks. He references his own company, Hebbia, noting that some users process 30 billion tokens in a single job, with line of sight to 100 billion-token jobs in 2026. The fifth pillar is outcomes. Sivulka states that most AI products today deliver cost savings, while enterprise leadership consistently prioritizes revenue growth. Institutional AI, he argues, must deliver measurable revenue upside rather than time savings alone.
Enablement, Unprompted Action, and the Road Ahead
The sixth pillar is enablement. Sivulka points to Palantir as an early example of a “process engineering” company, arguing that encoding firm processes into AI agents and managing organizational change will become the most important technology discipline in the near term. He notes that a top-three bulge bracket bank selected Hebbia for organization-wide deployment partly because competing model lab teams lacked domain knowledge.
The seventh and final pillar is unprompted action. Sivulka argues that requiring humans to prompt AI is a fundamental constraint, because humans rarely know the right questions to ask. He describes a scenario where an unprompted system monitors portfolio data continuously, detects deterioration in a company’s working capital cycle over three consecutive months, cross-references covenant thresholds, and alerts an operating partner before any human has reviewed the relevant document.
Sivulka concludes that individual AI and Institutional AI are complementary rather than competing. However, he states that organizations which adopt AI without redesigning their processes will lose ground to those that integrate technology and institutional structure together, repeating the pattern of the 1890s textile mills. The essay was published in the a16z newsletter on March 12, 2026, and had received 178 likes and 18 reposts at the time of publication. Sivulka is the founder of Hebbia, an AI platform focused on institutional knowledge work.