Silicon Valley is swollen to AI agents. Openai Sam Altman CEO said agents will “join the workforce” this year. Microsoft CEO of Microsoft Satya Nadella predicted that agents will replace certain knowledge tasks. Salesforce CEO Marc Benioff said that Salesforce’s goal is to be “The first digital work provider in the world“Through the various” agentic “services of the company.
But no one seems to agree on what an AI agent isexactly.
In recent years, the technology industry has boldly proclaimed that AI “agents” – the latest buzzword – are going to change everything. In the same way that AI Chatbots such as OpenAi Chatgpt have given us new ways for surface information, agents will fundamentally change the way we approach the work, we claim CEOs such as Altman and Nadella.
This may apply. But it also depends on how one defines “agents”, which is not an easy task. Like other terminology related to AI (eg “multimodal”, “agi” and “ai”), the terms “agent” and “agentic” are diluted to its point without meaning.
This threatens to leave Openai, Microsoft, SalesForce, Amazon, Google and countless other companies that build whole products around agents in an embarrassing place. An Amazon agent is not the same as a Google agent or any other seller, and this leads to confusion – and disappointment of customers.
Ryan Salva, senior director of the product on Google and a former Copilot leader, said he was “hating” the word “agents”.
“I think our industry goes beyond the term ‘agent’ to the point where it is almost stupid,” Salva told TechCrunch in an interview. “[It is] One of my pets. ”
The factor’s definition dilemma is not new. In a piece last year, former journalist TechCrunch Ron Miller asked: What is AI agent? The problem he recognized is that almost every corporate building agent approaches technology differently.
It is a problem that has recently aggravated.
This week, openai Posted a blog post that the specified factors as “automated systems that can achieve independent tasks on behalf of users”. However, the same week, the company was released Documentation of developers The defined factors as “LLMS equipped with instructions and tools”.
Leher Pathak, Openai’s API Product Marketing Lead, later stated in a Post in x That he understood the terms “assistants” and “agents” to be interchangeable – further muddy waters.
Meantime, Microsoft’s blogs try distinguish Between agents and assistants AI. The first, which Microsoft calls “New Applications” for a “world powered by AI”, can be adapted to have a specific experience, while assistants simply help in general duties, such as e -mail.
AI Lab Anthropic is a little more directly aimed at HodgePodge of agent definitions. To one blogAnthropic says agents “can be defined in various ways”, including both “fully autonomous systems that operate independently of extensive periods” and “prescription implementations following predefined work flows”.
SalesForce has this perhaps the most widespread definition of AI “Agent”. According to the software giant, Agents are ‘A guy […] The system that can understand and respond to customer surveys without human intervention. “The website of the company lists six different categories, ranging from” simple reflex agents “to” utility -based agents “.
So why the mess?
Well, agents – like AI – are a cloudy thing and are constantly evolving. Openai, Google and Purplexity have just begun sending what they consider to be their first agents – Openai’s operator, Google’s Mariner Project and Perplexity buyer – and their capabilities are all over the map.
Rich Villars, GVP of Worldwide Research at IDC, noted that technology companies “have a long history” not to follow the technical definitions rigidly.
“They take more care about what they are trying to achieve” on a technical level, Villars told TechCrunch, “especially in the rapidly evolving markets”.
But marketing is also largely responsible, according to Andrew Ng, the founder of the AI Deeplearning.ai learning platform.
“The concepts of” AI agents “and” practical “work flows used a technical meaning,” NG said in a recent interview, “but about a year ago, traders and some of their big companies took them.”
The lack of a unified definition for agents is both an opportunity and a challenge, says Jim Rowan, head of AI for Deloitte. On the one hand, ambiguity allows flexibility, allowing companies to adjust agents to their needs. On the other hand, it may – and undoubtedly has already – to lead to “abused expectations” and difficulties in measuring value and investment by projects.
“Without a standard definition, at least within an organization, it becomes difficult to refer to performance and ensure consistent results,” Rowan said. “This can lead to a variety of interpretations of what AI agents should offer. They may complicate the goals and results of the project. After all, while flexibility can lead to creative solutions, a more standardized understanding would help businesses better browse their AI landscape.”
Unfortunately, if the spread of the term “ai” is any indication, it seems unlikely that the industry will merge around a definition of “agent” at any time soon – if ever.
