How are AI Agents different from AI Chatbots?
This was a more technical question asked by Shayaan, a Senior Software Engineer:
𝘞𝘩𝘢𝘵 𝘪𝘴 𝘢𝘯 𝘈𝘐 𝘢𝘨𝘦𝘯𝘵? 𝘏𝘰𝘸 𝘪𝘴 𝘪𝘵 𝘥𝘪𝘧𝘧𝘦𝘳𝘦𝘯𝘵 𝘧𝘳𝘰𝘮 𝘊𝘩𝘢𝘵𝘎𝘗𝘛 𝘰𝘳 𝘊𝘭𝘢𝘶𝘥𝘦. 𝘢𝘪, 𝘦𝘷𝘦𝘯 𝘪𝘧 𝘸𝘦 𝘢𝘳𝘦 𝘶𝘴𝘪𝘯𝘨 𝘪𝘵𝘴 𝘈𝘗𝘐?
My answer:
AI agents are like a more advanced version of the chatbots we're used to.
We're already seeing them in action, for example, when you use a "deep research" tool. Instead of just answering a direct question like "How much do smartwatches cost?", a deep research agent would go beyond that.
It would perform additional searches to fully explore the underlying concerns, looking into things like smartwatch costs based on materials, where they're made, the economics of pricing, branding, and so on.
The key difference with an AI agent is that you give it a goal, and it figures out the best way to achieve it without you constantly giving more instructions.
More specifically, these agents can use tools outside of themselves, such as watching YouTube videos, Browse online marketplaces, reading reviews on sites like Yelp or Glassdoor, opening spreadsheets, and creating Word documents, all without you having to specifically tell them to.
In contrast, with APIs, I've only used them for very specific tasks.
Even then, I have to provide precise variables and instructions for them to work as intended.
That's not very "agent-like."