How I Actually Use Framer's AI Agent

What I delegate to the Agent, and what I keep

Framer 3.0 shipped with AI Agents, and most of the conversation since has been some version of the same question: Is this the thing that replaces designers?

I sell Framer templates for a living, so you'd expect me to be defensive about that. I'm not, because after using the Agent on real work for these couple of weeks, I've landed somewhere more boring and, I think, more useful than either the hype or the panic. The Agent didn't replace the part of my job I care about. It quietly took over the part I never wanted to do in the first place.

Here's the line I've drawn, and why.

The craft stays mine

Design is still the work. Deciding the typographic system, the spacing rhythm, where the eye lands first, how a section breathes, what gets cut, none of that gets delegated, because none of it is mechanical. It's judgment, and judgment is the whole product.

When I built Vitalo, my latest free template, the things that make it Vitalo were all decisions: pairing Playfair Display headings against Inter body text for that editorial-but-approachable feel, structuring 13 homepage sections so a personal trainer's site reads like a real business instead of a portfolio with photos swapped in. An Agent can't make those calls for you, and honestly I wouldn't want it to. That's the part I enjoy. That's the part customers are actually paying for.

So that's rule one: anything involving taste, composition, or craft stays in my hands, in the design tab, the same as it's always been.

The mechanical stuff goes to the Agent

Now the other half, the half nobody romanticizes.

A lot of my templates are agency and portfolio designs, and they ship with a Projects (or Work) CMS collection, the case-study grid a studio would fill with its own work. My design-niche demos, like Kudos and Xzero, both have one. And those project sections need to stay current: when I release something new, it should show up as fresh work inside the demos that are meant to showcase exactly that kind of design.

That used to be a manual chore, repeated across every demo. None of it is hard on its own, open a demo, add a new entry, paste the title and description, drag the image into place, set the fields, save. The problem is doing the exact same sequence again in the next demo, and the next, every time I release something. A few minutes of identical busywork per demo, per launch, and the kind of repetition where your attention drifts and you make exactly the small mistakes you're trying to avoid.

That's the job I now hand to the Agent.

When Vitalo went live, I created its entry once, properly, by hand, on my main site. That page became my single source of truth: uihub.design/templates/vitalo. Then, working in each design-niche template, Kudos, Xzero, I gave the Agent that one link and asked it to add a new project to the Projects collection: read the published page, pull the title, description, image, and key details, and fill the CMS fields with the relevant content.

It reads the source, understands the structure, and populates the new project entry. The repetitive part, recreating the same information by hand inside every demo, is now a single instruction.

That's the pattern, and it generalizes. The Agent is good at exactly the things design tools have always made tedious: bulk CMS entry, restructuring collections, repetitive edits across many pages, moving structured content from one place into another. Mechanical, rule-following, repetitive work with a clear right answer. Hand it that, and you get your time back for the work that actually needs you.

The part that makes it work: I still check it

Here's the caveat that keeps this honest, and it's not a footnote, it's the whole reason the workflow holds together.

I review everything the Agent creates before it goes live.

The fields don't always map identically between projects. An image crop that suits one layout might sit awkwardly in another. A description that fits one demo's grid might need a trim somewhere else. The Agent guesses well, but it does guess, and "well" isn't "always right." So the last step is always me: open the new entry, read it, fix anything that's off, then publish.

That review step isn't a weakness in the system, it is the system. It's the human judgment staying in the loop, just moved to the right place. I'm not building the new entry from scratch anymore; I'm approving and correcting one that's most of the way there. That's a completely different cost. But skip the check entirely and you'll eventually publish something off to a demo you forgot to look at, and the time you saved evaporates the moment someone spots it.

Where this leaves me

The framing that helped me most wasn't "AI versus designers." It was a simpler split: craft versus chores.

The craft, taste, structure, the decisions that make a template worth buying, stays human, because that's where the value lives. The chores, bulk CMS work, repetitive project entries, the same content recreated across multiple demos, go to the Agent, with me reviewing the output before anything ships.

That's not a compromise I feel bad about. I got into this to design, not to retype the same project description into a third CMS at 11pm. Framer 3.0 didn't come for the part of the job I love. It came for the part I'd have happily handed off years ago.

If you build or sell on Framer, I'd start there: not "what can the Agent do," but "what part of my week do I actively resent?" That's the part to delegate first. Keep the craft. Give away the chores. Check its work. That's the whole approach.

Vitalo is free, if you want to see the template this started with, grab it here.