Working with AI is starting to feel more like working with humans.
After hitting some limits and reconfiguring my AI setup, I now have a team of agents — each trained for specific tasks. One for planning, one for executing, one for writing, and even one for coding emails. Managing them and directing them isn’t far off from working with people. In the past week, I had the following interactions:
- Asking my engineer agent to review my design and share which use cases we’re missing and any technical barriers.
- Planning projects, breaking them down into smaller pieces, and sequencing them for my team of agents to execute.
- Guiding the agents when they’re blocked and spinning in circles.
- Running a retro at the end of the week to see what’s working and what could be better. Then making changes and applying fixes.
- Celebrating the wins and pointing out the failures (maybe sometimes a little too harshly :D).
It turns out 20 years of working with people prepared me for this moment more than I expected.