Agents are impressive. But let's talk about where they actually fit.

I've been playing with Framer Agents since 3.0 dropped and there's something I keep thinking about.

The demo looks great. Generate a layout, populate the CMS, fix responsiveness on mobile. Fast. Genuinely useful. But when I look at the outputs closely, I keep noticing the same thing: technically correct, creatively flat.

And I don't think that's a criticism of the technology. I think it's just an honest read of where we are right now.

Agents are exceptional at repetitive, rule-based tasks. CMS population, SEO fields, breakpoint fixes, component variants. Anything where the output can be evaluated against a clear standard. Hand that stuff off completely, it makes sense.

But design? The part where you decide that this layout needs more negative space, that this section should feel slower, that the hierarchy needs to shift because of how a specific client thinks about their brand? That's still a human judgment call. And if you outsource that to a prompt, the result tends to look like every other AI-generated site right now: structured but soulless.

I've been thinking about it as a co-pilot model, not an autopilot one. Agent does the scaffolding, I do the shaping. Agent handles the repetitive, I handle the intentional.

Curious if others are finding the same thing or if you've genuinely handed full design decisions to agents and been happy with where it landed.

Where are you drawing the line?

Agents are impressive. But let's talk about where they actually fit.

I've been playing with Framer Agents since 3.0 dropped and there's something I keep thinking about.

The demo looks great. Generate a layout, populate the CMS, fix responsiveness on mobile. Fast. Genuinely useful. But when I look at the outputs closely, I keep noticing the same thing: technically correct, creatively flat.

And I don't think that's a criticism of the technology. I think it's just an honest read of where we are right now.

Agents are exceptional at repetitive, rule-based tasks. CMS population, SEO fields, breakpoint fixes, component variants. Anything where the output can be evaluated against a clear standard. Hand that stuff off completely, it makes sense.

But design? The part where you decide that this layout needs more negative space, that this section should feel slower, that the hierarchy needs to shift because of how a specific client thinks about their brand? That's still a human judgment call. And if you outsource that to a prompt, the result tends to look like every other AI-generated site right now: structured but soulless.

I've been thinking about it as a co-pilot model, not an autopilot one. Agent does the scaffolding, I do the shaping. Agent handles the repetitive, I handle the intentional.

Curious if others are finding the same thing or if you've genuinely handed full design decisions to agents and been happy with where it landed.

Where are you drawing the line?