I’ve learned this hard truth from watching dozens of folks using AI agents.
The already productive ones get a 10X multiple on their productivity—but for a weird reason.
When you give an agent a task, it might not come back for 20-30 minutes.
So the “lazy” worker will goof off during that time, since they’re “working”.
I witnessed this phenomenon years ago with Filipino agents. A video editor, for example, would let the file render, then go play with their kids or have a meal for a couple hours—meanwhile, billing all that time.
The agents with slow internet connections would kick off a large file upload, giving them a few hours of free time.
When caught, their defense is that their computer or wifi is slow—that while the file is processing, they aren’t able to work. So we would buy two computers for agents.
Even still, people who aren’t driven will find excuses to not be productive.
If you’re like me, while a file is uploading, I move over to another tab, instead of just sitting there listening to my favorite music or going out to run some chores.
With AI agents, when I’m done giving instruction to one, I move on to the next one.
Admittedly, the context switching isn’t the most productive for me as a director. But the gains from having 10 agents running at the same time are undeniable.
You just have to be willing to play “whack-a-mole”, which means providing feedback as agents get stuck, need guidance on the next assignment, etc…
So the myth of AI agent productivity is limited not by the context window of the latest frontier model, but YOUR focus and the focus of your team members.
Related reading:
Honoring Shep Hyken: Repurposing Content with the Four-Stage Content Factory
