What do you all think about AI agents?
AI agents are trending all over the place on social media right now. Everyone’s talking about how capable they are. But personally, I think 90% of them don’t do much and perform poorly.
What’s your take on it?
Welcome to the forum
Post Guidelines
Please follow these guidelines for better discussions:
- Make sure your post is detailed, at least 100 characters.
- Search first. Many questions have already been answered.
- AI taking over jobs? That one’s been asked a lot!
- Feel free to talk about both good and bad aspects of AI, but keep it respectful.
- If you have a point, link to sources for support.
- No silly questions. We’re not here to talk about AI being the end of the world.
Let us know if you have any questions or comments.
I’m a bot, this post was auto-generated. Please reach out to the mods if you need help.
“90% of AI agents are useless and perform the task(s) poorly” – so, what’s the percentage for humans then?
I think AI is just like any tool. A great knife set is awesome, but it doesn’t mean you’re suddenly a chef. There are cases where AI agents are really helpful, but if you expect them to be all-knowing, you’re going to be disappointed.
Maybe I’m starting to agree with you. If the hype around AI agents is like those commercials convincing you that knives will make you a chef, then yeah, it sets up unrealistic expectations. But that doesn’t mean AI agents are totally useless, right?
It’s not as bad as it seems. With a little bit of research and input, you can teach your AI to perform tasks almost like a human. A lot of companies just aren’t training their agents properly.
They’re even better than some professionals…
Yeah, it’s a mix. Some of the hype is definitely deserved, but the way people sell AI agents is often misleading. Like when they said AI could write research papers, but if you actually read them, they don’t add anything valuable. Or when they claimed AI could replace programmers, but it still struggles with basic code.
It also bugs me how everyone is rushing to monetize AI agents like they’re going to replace humans. That’s not the case.
That said, big companies like Microsoft are doing cool things with AI. Copilot, for example, can help schedule appointments and respond to emails in their ecosystem. It’s not perfect, but it’s getting better.
To me, the real value of AI agents isn’t in the flashy use cases. It’s in smaller, less glamorous tasks that improve products behind the scenes. For example, an agent can extract key info from documents without a chatbot interface. That’s where AI should shine.
Source: I created the Atomic Agents framework, which focuses on minimalist AI that’s more useful for developers. Check it out here: https://github.com/BrainBlend-AI/atomic-agents. I also write a lot about AI agents on Medium, and I help companies implement them.
Don’t blame the AI for human mistakes. The issue is usually with the engineers who build the agents, or the users who don’t train them properly. AI isn’t perfect – you can’t expect it to work 100% from the start. It’s like expecting a child to write a thesis after learning the alphabet.
I think it’s good to start using AI agents now. They’ll get better as LLMs improve. But yeah, they’re a bit overhyped at the moment. Right now, it’s smarter to automate simpler tasks with reliable tools and use LLMs for the more complex stuff. A solid workflow that takes more effort to set up is better than an agent that’s quick to set up but inconsistent.
I agree! I’ve developed some that use 3.5Sonnet, and the inconsistency makes them almost useless, especially for people who aren’t familiar with coding.
Yeah, I think the hype around AI agents is definitely misguided in some cases.
Honestly, it’s usually faster and more accurate to just Google a few keywords than ask an AI. I keep seeing Gemini give wrong answers, even though a quick search leads to the right info in one click.
They’re saying 2025 is going to be the year for AI agents, so I’m waiting to see. Specifically OpenAI’s version called ‘Operator.’
Honestly, natural intelligence agents are way more overhyped.
We already know our limits, but it feels like AI agents still have a long way to go.