Silicon Valley is bullish on AI agents. Openai CEO Sam Altman says agency Will “join the workforce” this year. Microsoft CEO Satya Nadella predicts agency Will replace certain knowledge work. Salesforce CEO Marc Benioff said Salesforce’s goal is “The world’s number one digital labor provider“Through various “agent” services of the company.
But no one seems to have reached a consensus on AI agents yesexactly.
In the past few years, the tech industry has boldly announced that AI “agents” (the latest buzzword) will change everything. Just like an AI chatbot Openai’s Chatgpt CEOs like Altman and Nadella said that giving us new superficial information, agents will change the way we work.
Probably true. But it also depends on how people define “agents”, which is not easy. Like other AI-related jargons like “multi-mode,” “AGI,” and “AI” themselves), the terms “agent” and “agent” are downplayed to meaningless ideas.
This has the potential to leave Openai, Microsoft, Salesforce, Amazon, Google and countless others building their entire product lineup around agents in awkward places. Agents from Amazon are not the same as agents from Google or any other vendor, which leads to confusion – and customer frustration.
Ryan Salva, Google’s senior product director and former Github Copilot leader, said he began to “hate” the term “agent”.
“I think the term “agent” in our industry is almost ridiculous,” Salva said in accepting TechCrunch.[It is] One of my pets. ”
The proxy definition dilemma is not new. A piece from last year, former TechCrunch reporter Ron Miller ask:What is an AI proxy? The problem he identified is that almost every company construction agent treats technology in a different way.
This problem has worsened recently.
This week, Openai Post a blog post This agent defines the agent as an “automated system that can perform tasks independently on behalf of the user.” But in the same week, the company released Developer documentation This proxy defines the proxy as “LLM with instructions and tools”.
Leher Pathak, the head of API Product Marketing at Openai, later Post on X She understands that the words “assistant” and “agent” can be interchanged – further confusing the waters.
at the same time, Microsoft’s blog try distinguish Between the agent and the AI assistant. The former, which Microsoft calls the “AI-driven world,” calls it “new applications” that can be tailored to have specific expertise, while the assistant only helps with general tasks, such as drafting emails.
AI lab human human anthromorphization more directly solves the proxy-defined hodgepodge. exist Blog PostsAnthropic says agents can be defined in several ways, including “a fully autonomous system that runs independently for a long period of time” and “implemented by predefined workflows.”
Salesforce has the broadest definition of AI “proxy”. According to software giant Agent is “A type […] A system that can understand and respond to customer queries. “The company’s website lists six different categories, from “simple reflection agents” to “utilities-based agents.”
Then why be confused?
Well, agents like AI are a vague thing, they keep evolving. Openai, Google and Confused Just Begin to Ship They Think They Are Their First Agents – Operators at Openai,,,,, Google’s Project Sailorand Confused shopping agent -Their functions are spread throughout the map.
GVP Rich Villas of IDC Global Research noted that technology companies “hare a long history” and do not stick to the definition of technology.
“They care more about the work they are trying to accomplish,” Villars told TechCrunch, “especially in a fast-growing market.”
Andrew Ng, founder of AI Learning Platform, said marketing should also be largely blamed on Deeplelail.ai.
Ng said in a recent interview: “The concept of “AI’Adents’ and “agent” workflows have been technically meaningful in the past.
Jim Rowan, head of AI at Deloitte, said the lack of a unified definition of agents is both an opportunity and a challenge. On the one hand, ambiguity can increase flexibility, allowing companies to customize their agents according to their needs. On the other hand, it may and can be said to have resulted in “inappropriate expectations” and has encountered difficulties in measuring the value and return on investment of the agency project.
“There is no standardized definition, at least within the organization, benchmarking performance and ensuring consistent results become challenging,” Ron said. “This could lead to various explanations of what AI agents should offer and could complicate project goals and results. Ultimately, while flexibility can drive creative solutions, a more standardized understanding will help businesses better browse the AI agent landscape and maximize their investment.”
Unfortunately, if the revelation of the term “AI” is any indication, then if anything, it seems unlikely that the industry will soon merge around a definition of “agent.”
