Is Custome GPTs AI Agent?

Construction Industry Professionals Often Misunderstand AI Agents

I recently came across a post on my newsfeed introducing some AI tools for office documentation work targeting construction professionals. What caught my attention was the author's use of the term "AI Agent" to describe these AI tools. As someone who directly researches and applies AI in general and ChatGPT in particular in my work, I find this extremely problematic, as it will have serious consequences for foundational thinking during this early stage of AI development.

The Fundamental Misunderstanding

The tools the author mentioned are essentially Custom GPTs - a feature of ChatGPT. Simply put, these are ChatGPT instances enhanced with user-provided knowledge so that the AI (more precisely, LLMs - Large Language Models) can find and retrieve information more accurately according to requirements.

So how do Custom GPTs differ from an AI Agent? According to the definition (from ChatGPT), an Agent must both receive information and take actions based on the process of receiving and processing that information. While Custom GPTs do take actions (producing results), when most other AI platforms also have the action of producing text-based results, why do people need to "force" additional terminology unnecessarily?

The Growing Divide: Beginners vs. Professionals

From here, we begin to see differentiation in AI usage between beginners and professionals:

Beginners:

Input simple commands
Receive crude results
Follow "Ctrl C + Ctrl V" repetitive workflows
Work with isolated, disconnected AI types

Professionals:

Use true AI Agent systems where AI directly intervenes
AI "thinks through" implementation approaches on their behalf
AI directly executes multiple simultaneous tasks within permitted workspace and resources
No need for specific guidance - goal is final results and products
Eliminates intermediate steps

Real AI Agent Example

Scenario: You receive a notification that it's your business partner's birthday today. You're lying on My Khe Beach in Da Nang and don't want to open your laptop to write and send an email. So you command your AI Agent through your phone's chat box to investigate information about partner A, write a personalized birthday email, and send it, then put your phone down and continue sunbathing.

Ten minutes later, you receive a thank-you email from your partner saying they didn't expect you to know them so well and promising to visit soon - while you don't even know what content the AI wrote.

Current Market AI Agents

For AI to operate autonomously like this, there aren't many options currently available. Here are some notable AI Agents:

1. Devin AI

An "AI software engineer" developed by Cognition Labs
Independently plans, writes code, debugs, and even deploys applications
Latest version includes multi-agent capabilities and self-assessment of confidence levels

2. Manus

Comprehensive agent considered to have true autonomous capabilities
Developed by startup Monica.im in China
Officially launched in March 2025
Can plan, automatically write and deploy code, operate efficiently without direct guidance

3. ChatGPT Agent

Officially launched by OpenAI on July 17, 2025
Built-in system within ChatGPT combining real-world action capabilities (Operator) and deep research
Uses a "virtual computer" where ChatGPT can flexibly switch between tools like browser, terminal, code execution, APIs, and connectivity tools like Gmail or GitHub

4. Project Mariner

Research prototype from Google DeepMind
Designed to intelligently interact with web browsers
AI agent running in Chrome that can "see," interpret, and execute web tasks like humans: clicking, scrolling, filling forms, finding information, etc.

5. Perplexity Comet – "Agentic" AI Browser

Dedicated web browser developed by Perplexity
Specifically designed to operate as an AI Agent within the browser
Users can request Comet to perform tasks in natural language: browsing, summarizing content, filling forms, scheduling, shopping, sending emails... all done directly within the browser

The Challenge Ahead

While convenient, the emergence of AI Agents invisibly creates new challenges related to data security and AI ethics. And these are issues we still don't have solutions for.

Which LLMs that you are using?

ChatGPTs

Claude AI

Gemini

Grok

Perplexity

DeepSeek

2 votes

0 comments