AI Agent Definition Eludes Even Top Venture Capitalists

Tech buzzwords often lose meaning through overuse, and "AI agent" is the latest culprit, with terms like "agentic" adding to the confusion.

Surprisingly, even experts at Andreessen Horowitz, a leading venture capital firm heavily investing in AI startups, admit there's no clear consensus on what an AI agent is.

In a recent podcast episode titled "What Is an AI Agent?", three a16z infrastructure investment partners—Guido Appenzeller, Matt Bornstein, and Yoko Li—attempted to define the term.

For context, a16z, which backs prominent AI firms like OpenAI and Anysphere (creator of Cursor), is reportedly seeking a $20 billion fund to further fuel AI investments, according to Reuters. In a September blog post, two a16z VCs expressed enthusiasm, stating: "We envision AI copilots for every white-collar role, with some fully automated by AI agents."

Appenzeller notes that many AI startups are capitalizing on the hype by labeling their products as agents, creating a broad spectrum of interpretations.

"The simplest so-called agent is just a smart prompt layered over a knowledge base," Appenzeller explained. Such systems take a human query and deliver a pre-set response, like those used in IT help desks.

Recently, however, companies have begun marketing AI agents as replacements for human workers.

Showcase at TechCrunch Sessions: AI

Claim your spot at TC Sessions: AI to present your innovation to over 1,200 decision-makers without breaking the bank. Available until May 9 or while spots remain.

Showcase at TechCrunch Sessions: AI

Claim your spot at TC Sessions: AI to present your innovation to over 1,200 decision-makers without breaking the bank. Available until May 9 or while spots remain.

Appenzeller argues that true human replacement would require AI approaching general intelligence, capable of persistent operation and independent problem-solving.

"That technology isn’t there yet," both he and Li confirmed.

Jaspar Carmichael-Jack, CEO of Artisan, an AI sales agent startup, told TechCrunch that perfecting reliable AI agent technology remains a tough challenge. Despite Artisan’s viral "stop hiring humans" campaign, Carmichael-Jack continues to employ human workers.

Significant hurdles, like enabling long-term memory and eliminating AI hallucinations, must be overcome for AI to truly replace human workers. No company wants an agent—human or artificial—that forgets prior interactions or provides false information.

In the podcast, the a16z trio settled on a practical definition: an AI agent is a multi-step, reasoning language model with a dynamic decision-making framework.

Li clarified that an agent goes beyond task execution; it must independently decide and act, such as retrieving prospect lists from a database, selecting email recipients, and drafting messages—or writing and placing code autonomously.

On the potential for AI agents to replace humans, the VCs agreed that agents could handle certain tasks, much like past automation. However, this might increase demand for human workers as productivity grows, rather than reduce it.

Bornstein emphasized that human creativity and critical thinking remain irreplaceable. From Silicon Valley’s perspective, he noted, it’s easy to overlook that most jobs rely on uniquely human skills, making full replacement theoretically questionable.

Bornstein added that exaggerated claims about human replacement, often driven by marketing or pricing strategies, contribute significantly to the current confusion around AI agents.

Ultimately, if even those at the forefront of AI innovation question the grand promises of AI agent companies, it’s a cue for everyone to approach such claims with skepticism.