No one knows what the hell an AI agent is | TechCrunch

Silicon Valley is bullish on AI agents. OpenAI CEO Sam Altman said agents will “join the workforce” this year. Microsoft CEO Satya Nadella predicted that agents will replace certain knowledge work. Salesforce CEO Marc Benioff said that Salesforce’s goal is to be “the number one provider of digital labor in the world” via the company’s various “agentic” services.

But no one can seem to agree on what an AI agent is, exactly.

In the last few years, the tech industry has boldly proclaimed that AI “agents” — the latest buzzword — are going to change everything. In the same way that AI chatbots like OpenAI’s ChatGPT gave us new ways to surface information, agents will fundamentally change how we approach work, claim CEOs like Altman and Nadella.

That may be true. But it also depends on how one defines “agents,” which is no easy task. Much like other AI-related jargon (e.g. “multimodal,” “AGI,” and “AI” itself), the terms “agent” and “agentic” are becoming diluted to the point of meaninglessness.

That threatens to leave OpenAI, Microsoft, Salesforce, Amazon, Google, and the countless other companies building entire product lineups around agents in an awkward place. An agent from Amazon isn’t the same as an agent from Google or any other vendor, and that’s leading to confusion — and customer frustration.

Ryan Salva, senior director of product at Google and an ex-GitHub Copilot leader, said he’s come to “hate” the word “agents.”

“I think that our industry overuses the term ‘agent’ to the point where it is almost nonsensical,” Salva told TechCrunch in an interview. “[It is] one of my pet peeves.”

The agent definition dilemma isn’t new. In a piece last year, former TechCrunch reporter Ron Miller asked: What’s an AI agent? The problem he identified is that nearly every company building agents approaches the tech differently.

It’s a problem that’s worsened recently.

This week, OpenAI published a blog post that defined agents as “automated systems that can independently accomplish tasks on behalf of users.” Yet in the same week, the company released developer documentation that defined agents as “LLMs equipped with instructions and tools.”

Leher Pathak, OpenAI’s API product marketing lead, later said in a post on X that she understood the terms “assistants” and “agents” to be interchangeable — further muddying the waters.

Meanwhile, Microsoft’s blogs try to distinguish between agents and AI assistants. The former, which Microsoft calls the “new apps” for an “AI-powered world,” can be tailored to have a particular expertise, while assistants merely help with general tasks, like drafting emails.

AI lab Anthropic addresses the hodgepodge of agent definitions a little more directly. In a blog post, Anthropic says that agents “can be defined in several ways,” including both “fully autonomous systems that operate independently over extended periods” and “prescriptive implementations that follow predefined workflows.”

Salesforce has what’s perhaps the most wide-ranging definition of AI “agent.” According to the software giant, agents are “a type of […] system that can understand and respond to customer inquiries without human intervention.” The company’s website lists six different categories, ranging from “simple reflex agents” to “utility-based agents.”

So why the chaos?

Well, agents — like AI — are a nebulous thing, and they’re constantly evolving. OpenAI, Google, and Perplexity have just started shipping what they consider to be their first agents — OpenAI’s Operator, Google’s Project Mariner, and Perplexity’s shopping agent — and their capabilities are all over the map.

Rich Villars, GVP of worldwide research at IDC, noted that tech companies “have a long history” of not rigidly adhering to technical definitions.

“They care more about what they are trying to accomplish” on a technical level, Villars told TechCrunch, “especially in fast-evolving markets.”

But marketing is also to blame in large part, according to Andrew Ng, the founder of AI learning platform DeepLearning.ai.

“The concepts of AI ‘agents’ and ‘agentic’ workflows used to have a technical meaning,” Ng said in a recent interview, “but about a year ago, marketers and a few big companies got a hold of them.”

The lack of a unified definition for agents is both an opportunity and a challenge, Jim Rowan, head of AI for Deloitte, says. On the one hand, the ambiguity allows for flexibility, letting companies customize agents to their needs. On the other, it may — and arguably already has — lead to “misaligned expectations” and difficulties in measuring the value and ROI from agentic projects.

“Without a standardized definition, at least within an organization, it becomes challenging to benchmark performance and ensure consistent outcomes,” Rowan said. “This can result in varied interpretations of what AI agents should deliver, potentially complicating project goals and results. Ultimately, while the flexibility can drive creative solutions, a more standardized understanding would help enterprises better navigate the AI agent landscape and maximize their investments.”

Unfortunately, if the unraveling of the term “AI” is any indication, it seems unlikely the industry will coalesce around one definition of “agent” anytime soon — if ever.