I write code, I run marketing, and most of my work is in the gap between them.
I started as a developer, founded and sold a Webflow agency, and now lead marketing across a network of agencies, and I still build: websites, marketing systems, AI workflows. That sequence is less a career path than an argument. It is the reason I can follow a technical change from the commit through to what it does to a marketing team on Monday morning.
What I work on
Most of my work is translation. A technical shift arrives, and a marketing leader has five questions about it that almost nobody answers well.
- 1
What the technology actually makes possible. Not the demo. The thing that changes for a team of eight people who already have a roadmap.
- 2
What it does not solve by itself. Every new capability quietly assumes you already fixed a set of problems you have not fixed.
- 3
How workflows and teams have to change. New capability dropped into an old workflow produces faster versions of the same output, which is not the same as better output.
- 4
What context, systems, and governance have to exist first. Positioning, a design system, a content model, a quality bar, and a named owner for each. AI is only as good as the definitions you can hand it.
- 5
How capability becomes execution. The distance between "we could do this" and "we do this every week" is organizational, not technical, and it is where most of the value sits.
Where this comes from
I have built, sold, run, and shipped, which is a different vantage point than commenting from the outside. In 2018 I founded Belt Creative, a Webflow design and development agency, grew it to a team of 10 and more than $1M in annual revenue building sites for mid-market B2B companies, and sold it to Veza Digital in October 2025. Today I am CMO of Veza Agency Network, which I treat as current proof rather than the name on the door: it is where I run into the problems I write about, at a scale where they cannot be hand-waved. I am still in the CMS, still in the repository, and still in the room where the design gets argued about.
Ideas I'm developing
Four threads I am actively working out, in public, against work I am actually doing.
The marketing operating system
Marketing tools are organized around artifacts. Marketing work is organized around intent. "Launch a webinar" is one thing to a leader and six unrelated assignments to the stack: Figma, the CMS, HubSpot, analytics, and a spreadsheet to track the other four. I think the resolution is a layer that holds the intent and the context and drives the tools, and I think the website is the natural home for it, because the website is the only artifact where positioning, design system, components, content, and measurement have already been forced to agree. Webflow Source is a useful lens on this. It is not the point.
Where this is going: I am writing the case for the website as the system of record, and looking for the version of it that survives contact with a real team.
Humans design the system. AI operates within it.
The useful boundary is not hard work versus easy work. It is who defines what correct means and who executes inside a definition that already exists. Humans own positioning, the design system, the quality bar, taste, and the decision to publish. AI works inside those, assembles from approved templates, produces contextual variations, and carries context from step one to step nine. The uncomfortable part is that the second half only works if the first half was written down, and in most organizations it never has been.
Where this is going: I am collecting the failure cases, because the split is easy to state and hard to hold.
The human integration layer
At Veza Agency Network I am the person who knows where the assets are, which repository is current, which prompt produces acceptable copy, and whether a draft is on brand. That looks like competence. It is a bottleneck. Our marketing context is fragmented across Sanity, GitHub repositories, prompts, skills, asset folders, individual tools, and people's heads, and I am the part that reconciles them, which caps throughput at my availability and makes the output inconsistent whenever I am not consulted.
Where this is going: I am trying to write myself out of the role, and documenting what has to exist before that is possible.
Translation, not evangelism
There is no shortage of people telling marketing leaders that AI changes everything. There is very little writing about what a specific technical change means for a specific marketing organization: which roles change, which workflows break, what has to be built before any of it works, and what to do about it this quarter. I am more interested in the second thing, partly because it is harder and mostly because it is checkable.
Where this is going: every idea here gets tested against work I am doing, and I will say so when one of them stops holding up.
Writing
Long-form thinking on four threads: the marketing operating system, AI in practice, building and selling, and translating technical change. Ideas move here once they hold up under an argument.
There is nothing published here yet. The first essays are written and in review. Until they are up, the four ideas above are the outline, and X is where I work through them in the open.