‎

I’ve definitely noticed this shift. I used to care mostly about how many features a SaaS product had, but now I care more about whether it can actually help me finish the task. If I still have to guide the AI through every little step, the “agent” part doesn’t feel very useful.

‎

‎

‎There’s also a maintenance angle here. One huge agent handling everything sounds convenient until you’re debugging why it made a bad decision. Smaller agents with clear responsibilities could be easier to test and improve.
‎

‎I haven’t used Backstage enough to judge it yet, but I like the direction. AI managing the AI agent itself feels like a pretty natural next step. The real test for me would be whether it understands the context of the agent instead of just acting like another settings shortcut.
‎