Owner-Operator: Running a P&L with an AI-Native Marketing Stack

The situation

In August 2024 my wife and I became owner-operators of Upgrade Labs Park City, a brick-and-mortar health and human performance business. Real lease, real staff, real customers walking through a real door, and a P&L that lands on our kitchen table, not in someone else’s board deck.

I include this alongside the fintech case studies deliberately. Executives at my level usually describe systems their teams ran. This is the proof that I still personally ship: every system described below, I built or run hands-on, today.

What I walked into

A local services business in a category, recovery and human performance, where customer acquisition is won or lost in hours, not quarters. A lead who fills out a form and hears nothing for a day books somewhere else or nowhere at all. There was no growth engine to inherit. Like Nav in 2015, but with a compressed timeline and my own capital, it had to be built from zero, and it had to work without a team of twenty behind it.

The constraint is the interesting part. A single-location P&L cannot afford an agency retainer for every function, which makes it the perfect forcing function for a question I had been circling for years: how much of a modern marketing organization can one operator run with AI agents carrying the heavy work?

The moves

1. Make Meta lead-form campaigns the primary acquisition channel, run hands-on. I build the campaigns, the audiences, the offers, and the creative myself. Lead-form ads fit a local services funnel because they collapse the distance between interest and contact info, and because cost per lead is a number you can manage daily against a P&L you can feel.

2. Automate speed-to-lead. Every lead gets an SMS follow-up in minutes, automatically, with multi-day nurture sequences behind it for the ones who do not book immediately. In local services, response time is the highest-return variable in the funnel, and automation is the only way a small team wins it consistently.

3. Build the operations layer with AI agents. The reporting command center, the financial forecasts, the commission calculations, and a growing share of ad creative are built and maintained by AI agents I direct. Work that would have been a fractional analyst, a bookkeeper’s spreadsheet, and a design retainer is a set of agent workflows I can rebuild in an afternoon.

4. Add clinical services through partnership rather than build. GLP-1, peptide, and hormone therapy are delivered in partnership with OutfitMD. This is a claims-disciplined category: the marketing has to stay on the right side of health-claims rules, which means the copy discipline matters as much as the channel strategy.

5. Run the whole P&L. Pricing, staffing, lease economics, cash flow, marketing spend. Not a budget owner inside someone else’s model. The model.

The numbers

• P&L ownershipFullpricing, staffing, cash, spend
• Primary channelMetalead-form campaigns, run hands-on
• Speed-to-leadMinutesautomated SMS and multi-day nurture
• Ops built byAI agentsdashboards, forecasts, commissions, creative

This business’s revenue figures stay private. The systems are the story.

What broke along the way

Brick-and-mortar punishes assumptions that survive fine in software. Demand in a mountain resort town is seasonal in ways a SaaS dashboard never prepared me for, and a slow week is not a chart dipping, it is payroll against a quiet room.

Automation broke in the places automation always breaks: the edges. A nurture sequence that reads perfectly in a planning doc will eventually message someone at the wrong moment, and in a small town your brand is a neighbor, not an impression. The sequences needed guardrails, suppression logic, and a human override before they earned trust.

And the honest AI lesson: agents amplify the operator, they do not replace judgment. An agent-built forecast is only as good as the assumptions I feed it, and I have had to throw out and rebuild agent workflows that produced confident, tidy, wrong answers. The skill is knowing what to check.

What I would tell you if you are facing this

Every marketer should run a P&L they can feel at least once. It rewires how you spend other people’s money forever. The empathy I now have for the operators my fintech career served, small business owners making payroll on Friday, came from standing where they stand.

If you are an executive wondering whether the AI stack talk is real, here is the test I would give any candidate, because I can pass it: ask them to show you a system they personally built with AI that runs their business today. Not a pilot, not a deck. A dashboard, a forecast, a sequence, in production.

And in a claims-sensitive category, treat compliance as a copy skill. The operators who thrive next to health claims rules are the ones who learn to sell outcomes-adjacent benefits precisely, not the ones who flirt with the line.

More of the record: all case studies · the writing · get in touch