The concept of buzzwords losing their meaning is not new to the tech industry, and currently, the term “AI agent” and its variants, such as “agentic,” are being used loosely. As a result, there is a lack of clarity on what exactly an AI agent is, even among software engineering professionals at prominent venture capital firms like Andreessen Horowitz.

A recent podcast episode, “What Is an AI Agent?”, featured three a16z infrastructure investment partners – Guido Appenzeller, Matt Bornstein, and Yoko Li – attempting to define the term. Despite their efforts, it became clear that there is no universally accepted definition of an AI agent, even among experts in the field.

Andreessen Horowitz, a backer of popular AI companies like OpenAI and Anysphere, is heavily invested in the AI sector and is reportedly seeking to raise a $20 billion mega-fund to further invest in AI startups. The firm’s enthusiasm for AI is evident, with two of its VCs stating that every white-collar role will have an AI copilot, and some roles will be fully automated with AI agents.

However, the term “AI agent” has become a buzzword, with many startups describing their products as agents to capitalize on the trend. Appenzeller notes that some companies are calling simple AI-powered tools “agents,” which can perform tasks like fetching canned responses to user queries.

Some AI startups are now claiming that their agents can replace human workers, but experts argue that this would require significant advancements in AI technology, including the ability to persist over long periods and work independently on problems. Currently, such capabilities do not exist, and even the most advanced AI agents are not reliable enough to replace human workers.

The CEO of sales AI agent company Artisan, Jaspar Carmichael-Jack, acknowledged the challenges of developing reliable AI agents, stating that the journey has been harder than expected. Despite his company’s “stop hiring humans” ad campaign, Carmichael-Jack is still hiring human workers.

To create an AI agent that can truly replace human workers, significant technical hurdles need to be overcome, such as developing persistent long-term memory and eliminating hallucinations. These challenges are significant, and no company wants to hire an AI agent that cannot remember previous conversations or provides inaccurate information.

During the podcast, the a16z trio did manage to define what is possible with current AI technology. Li described an AI agent as a reasoning, multi-step LLM with a dynamic decision tree, which can make decisions and take action autonomously.

In essence, an AI agent is not just a bot that performs a task when asked, but one that can also decide on the task and take action independently. This can include tasks like writing code, deciding where to insert it, or grabbing a list of prospects from a database and sending emails.

While AI agents may be able to handle some tasks currently performed by humans, the a16z VCs agreed that this may lead to companies hiring more human workers, not fewer, as productivity increases. Bornstein stated that he cannot envision a time when humans will be unnecessary, given the current state of AI agents.

The notion that AI agents can replace humans is often used as a marketing tool, which has contributed to the confusion surrounding the term. Bornstein argues that this rhetoric is not only misleading but also forgets that most people have jobs that require human creativity and thinking.

In conclusion, the skepticism surrounding the claims made by AI agent companies is justified, and it is essential to approach these claims with a critical eye. The current state of AI technology is not yet advanced enough to replace human workers, and significant technical challenges need to be overcome before AI agents can become a reality.