OpenAI’s AI agents have around 10 million weekly users. That sounds impressive, right? Now compare it to ChatGPT’s roughly one billion monthly users. AI agents barely register. The gap between what Silicon Valley is building and what regular people actually use is massive, and it tells you something important about the state of AI right now.
Tech companies have spent billions on AI agents. You’ve probably seen the demos: an AI that books your flights, manages your inbox, or builds entire apps from a sentence. Cool stuff. But most people have never touched one. That’s not an accident.
The weird gap between AI hype and actual use
Josh Miller, CEO of The Browser Company, recently posted something that struck a nerve in Silicon Valley. “Hot take,” he wrote, “isn’t it kinda crazy that nobody is really using AI agents.” His point was simple: the technology exists, the demos are impressive, and the general public doesn’t care.
He’s right. And the reasons aren’t what you’d expect.
The numbers don’t lie
ChatGPT has around a billion monthly active users. Gemini is in the same ballpark. These are chatbots that answer questions, write emails, and summarize documents. Straightforward stuff.
AI agents, which are supposed to do far more (execute tasks, make decisions, interact with multiple apps), have roughly 10 million weekly users across OpenAI’s platforms. That includes Codex, ChatGPT Work, and all the agent features combined. Anthropic’s Claude Code and Cowork agents are reportedly at similar levels.
For context: AI agents represent less than 1% of total AI tool usage. Billions invested, and almost nobody shows up.
Why most people avoid AI agents (the real reasons)
This isn’t about people being “slow to adopt.” There are specific, fixable reasons why AI agents haven’t broken through. And understanding them matters whether you’re building these tools or just trying to figure out which ones are worth your time.
They are too complicated to set up
Most AI agents require some level of technical setup. Connecting APIs, configuring permissions, understanding authentication flows. For someone who just wants their inbox sorted, this is a non-starter.
The average person’s tolerance for setup before seeing value is about two minutes. If an AI agent takes 15 minutes to configure, most users will bounce before it does anything useful.
Compare that to ChatGPT, which works the second you open the page. No setup. No API keys. Just type. That friction difference explains a lot of the adoption gap.
Nobody explains what they actually do
Search “AI agent” and you’ll find a hundred articles about agentic frameworks, recursive self-improvement, and multi-step reasoning. Very few explain what these things do in plain English.
“An AI agent is a system that can autonomously plan and execute multi-step tasks using tools and APIs.” Great. Now imagine explaining that to your mom. The marketing is written for developers, not the people who’d actually benefit from using these tools.
Here’s what most people want to hear: “This thing can book your dentist appointment, sort your emails, and remind you to pay your electricity bill. Here’s how to turn it on.” That’s it. The rest is noise.
Trust issues are real
People are wary of giving an AI access to their email, calendar, bank accounts, or workplace tools. And honestly, they should be. The same week OpenAI announced its agent features, security researchers were showing how AI agents could hijack browsers and make unauthorized purchases.
Until AI agents earn trust through reliability and transparency, most users will keep them at arm’s length. That’s not stubbornness. That’s common sense.
Chatbots already feel like enough
For most people, asking ChatGPT or Gemini a question and getting an answer in seconds is already magic. The jump from “AI answers my questions” to “AI autonomously executes tasks across my apps” doesn’t feel necessary to the average user.
“I can already ask ChatGPT to draft an email. Why do I need an agent to send it?” This is a real sentiment, and it’s hard to argue with. The value proposition of agents needs to be obvious and immediate, not theoretical.
AI agents that are actually worth trying right now
Despite all of this, there are some AI agents that have crossed the “actually useful” threshold for regular people. These are the ones worth trying if you’re curious but haven’t taken the plunge.
ChatGPT Tasks — the agent you probably already have
If you use ChatGPT, you might already have access to Tasks. It’s OpenAI’s take on AI agents, and it’s built directly into the chat interface. You can ask it to do things like monitor web pages, send reminders, or compile daily briefings. The setup is minimal because it lives inside a tool you already use.
Not mind-blowing, but practical. It’s the kind of agent feature that doesn’t require a tutorial. Check out our full guide to ChatGPT Work for more on what OpenAI’s agent tools can do.
Claude Code — for building things without coding
Claude Code from Anthropic is a terminal-based agent that writes code, builds apps, and automates tasks. You don’t need to know how to code. You just describe what you want in plain English, and Claude Code does the rest.
It’s found an audience among non-technical users who want to build simple tools, automate workflows, or create websites. We covered how to use Claude Code for 3D printing without any CAD experience, and the principle is the same for other tasks.
AI scheduling agents
Scheduling is one area where AI agents genuinely save time. Tools like ChatGPT’s Schedule feature and dedicated scheduling assistants can coordinate meetings, find time slots across calendars, and handle the back-and-forth that normally eats up your morning. We recently covered the best AI scheduling assistants that do this well.
The bottom line on AI agents
AI agents aren’t failing because the technology isn’t ready. They’re failing because the industry built tools for itself instead of for everyone else. Too much setup, too much jargon, too little trust.
That’s changing, slowly. The agents that are gaining traction are the ones that work inside existing tools, require zero configuration, and deliver obvious value within seconds. If you’re curious about AI agents, start with something simple. ChatGPT Tasks or an AI scheduling assistant are good first steps.
For a deeper look at how agents work, our explainer on AI browser agents breaks down what they do and which ones are worth trying. And if you want to connect agents to your own apps, our step-by-step guide walks you through the process for free.
