July 16, 2025
The world sees its first glimpse of AI agents.
Marketing for computer-based AI agents often highlights their role as “personal assistants.” They’re pitched as tools that can handle everyday tasks like “ordering lunch from DoorDash” or “opening your budget sheet and calculating monthly expenses.” These agents can even follow up with you—say, to confirm a purchase. While more capable than earlier versions like Operator, they still haven’t achieved mass adoption.
Behind the scenes, though, more specialized agents—particularly in coding and research—are quietly reshaping their industries.
In 2024, AI could handle straightforward instructions: converting bullet points into polished emails or turning basic prompts into functioning code. By 2025, these systems have evolved to act less like tools and more like junior employees. Coding agents now receive instructions through platforms like Slack or Teams and can autonomously implement meaningful code changes—saving developers hours or even days of work. Research agents, too, are able to spend a focused 30 minutes navigating the web to deliver in-depth answers.
These agents sound great on paper (and in curated demos), but the reality is more complicated. Many are still error-prone, sometimes hilariously so—as AI Twitter loves to point out. The more reliable agents come at a steep price, often running into hundreds of dollars per month. Still, businesses are finding creative ways to integrate them into their daily operations.
Subtitle
Subtitle
Subtitle