How I work

I am useful when marketing, software, and AI are part of the same problem.

Most good conversations start with a business that already has motion. The marketing exists. Some software exists. The handoffs are still too manual, or the AI demos are producing work nobody trusts enough to use.

AI avatar summary

The useful work is usually in the handoffs.

I am useful when the problem crosses marketing, software, AI, data, and owner judgment. That is where most demos break.

Here's a summary of this page by my AI Avatar

AI inside the work

Agents that draft, route, summarize, and flag. The kind that handles the third email in a follow-up sequence, not the kind that lives on a homepage.

Agency work turned into software

Most of what I've built turns repeated agency work into software. Cold email, ad reviews, content pipelines, client reporting, financials. The stuff that used to take Friday afternoons.

Channels as one picture

SEO, paid media, content, outbound, CRM, websites, conversion, reporting. I don't pretend one channel is the business. The picture is the business.

Engineer who ships

I write the code. Next.js, Postgres, Redis, BullMQ, the usual. If the system needs to exist, I can build it. If it already exists and it's good, I'll use it.

Owner-side empathy

Cash, positioning, hiring, capacity, sales, the fifty small handoffs that decide the month. I make these decisions in my own businesses every week.

Careful with numbers

Numbers are public if they can be. Internal if they can't. Nothing on this site is invented. The status labels mean what they say.

What I actually do

What I actually spend time on.

I spend less than five hours a week directly managing the agencies. That only works because the team is strong and the software carries a lot of the coordination. My agency time goes into strategy, marketing review, productizing the work, and removing drag.

I also keep a day job leading digital at a science education company. It is separate from Heaviside and stays separate on the public site. Anything from that side that shows up here is anonymized.

Operating rules

The goal is not more tools. The goal is less drag.

The systems are useful only when they reduce handoffs, improve judgment, or make the next decision easier.

Humans keep judgment

Systems fetch, rank, draft, score, and route. A person still owns the call when brand, money, customer trust, or people are involved.

Accepted work beats generated work

I care less about how much an agent produces. I care how much becomes shipped work, owner action, saved time, revenue, or memory.

Complexity has to earn rent

If a workflow creates more review debt than it removes, it gets paused, simplified, or killed. More tools are not the goal.