Follow ZDNET: Add us as a favorite source On Google.
ZDNET Highlights
- MIT released a list of top AI agents and their functionalities.
- The largest portion focuses on enterprise workflows.
- Research and information synthesis are the top use cases.
Which autonomous or semi-autonomous agents are having the biggest impact on the world – and possibly your job – these days? Some agents have been grabbing all the headlines lately, but there are many function-specific agents available to developers and users.
from mit C.S.A.L – a university laboratory dedicated to AI research – designed to identify and document the backgrounds and capabilities of these agents, detailing its latest findings. AI Agent Index. The researchers conducted an ecosystem-wide analysis of state-of-the-art AI agents on 1,350 data points.
Also: MIT study shows AI agents are fast, loose, and out of control
What is the functionality and origin of leading agents? The researchers found that interfaces were the most abundant, followed by enterprise workflow platforms. They also highlight the risks shared by these agents, as my ZDNET colleague Tiernan Ray discovered.
Agents featured in the MIT Index include the following:
- Anthropic Cloud/Cloud Code
- Google Gemini/Gemini CLI
- Manus AI
- OpenAI ChatGPT/ChatGPT Agent/Codex/AgentKit
- distress
- Alibaba MobileAgent
- ByteDance Agent TARS
- distraction comet
- IBM WatsonX Orchestra
- Microsoft 365 Copilot
- SAP Joule Studio
- salesforce agentforce
- ServiceNow AI Agent
Here are the three major categories of agents identified by researchers:
Enterprise Workflow Agent (contains 13 of 30 systems): These are platforms with agentic features to automate business tasks. Examples include Microsoft 365 Copilot and ServiceNow Agent.
Chat application with agentic tools (12 systems): According to researchers, this category mainly includes chat interfaces with extensive tool access. Examples include general-purpose coding agents such as Cloud Code, as well as agents embedded in broader products such as Manus AI and the ChatGPT agent.
Plus: I gave Chrome’s AI agent permission to shop, research, and email for me — here’s how it happened
Browser-based agents (5 systems): These are agents whose primary interface is using a browser or computer, including extensive browser/computer interaction tools. “They are different from chat agents with web search capabilities – ChatGPT Web Search, Cloud Web Search – which primarily perform retrieval and summarization,” the researchers say. “Browser-based agents present higher risks through background execution, event triggers, and direct transactions.” Examples include Perplexity Comet, ChatGPT Atlas, ByteDance Agent TARS.
What are the most popular uses for AI agents?
The top use cases for AI agents in the above categories include research and information synthesis, as seen in 12 of the 30 agents covered, including both consumer chat assistants and enterprise platforms. Right behind this functionality is workflow automation across business functions – such as HR, sales, support and IT – enabled by 11 agents, found primarily in enterprise products. Agents focused on GUI or browser capabilities, used for tasks such as forms, ordering, and booking, exist in seven models.
Also: 6 reasons why autonomous enterprises are still more a fantasy than a reality
The researchers found that the level of autonomy varies considerably. Chat-first assistants maintain the lowest level of autonomy. These are based on turn-based interactions, and include Anthropic Cloud, Google Gemini, and OpenAI ChatGPT, which “performs a set of tasks and waits for the next user prompt.”
At higher levels of autonomy, browser agents provide “more limited opportunities for mid-execution intervention”. These include Perplexity’s Comet, which acts autonomously when signaled. “Once a query is sent, users cannot easily interfere with it or control the agent until it is finished.”
Enterprise platforms are divided when it comes to agent autonomy. “During the design phase, users manually configure triggers, actions, and guardrails using a visual canvas,” the researchers wrote. Others may offer AI assistance in the process. Once deployed, these agents often operate at a high level of autonomy, “triggered by events such as a new email or database change, without any human involvement during actual task execution.” Such agents include Glenn, Google Gemini Enterprise, IBM WatsonX, Microsoft 365 Copilot, N8N, and OpenAI AgentKit.
Also: Google Workspaces now lets you create AI agents to automate your work – how to get started
Some offerings are considered developer/command-line-interface (CLI) agents that require explicit confirmation for sensitive operations such as file editing and command execution. Some agents offer “watch mode” for real-time monitoring of critical functions, including ChatGPT Agent/Atlas and Opera Neon.
The study also found that agent developers are concentrated in the US and China, with limited representation from other regions.
