Operating System • How I Work • v1.0 — April 2026
How I Work With AI as a Performance Marketer
(And How to Work With Me)
If you’re working with me — or trying to replicate how I think — this is the operating system.
Why this exists
Marketing should not be a guessing game. It should be measurable, accountable, and tied directly to business outcomes.
This document explains how I approach:
And how to get the most out of working with me (or thinking like me).
My Core Philosophy
1. Marketing must tie to revenue
If we can’t connect what we’re doing to leads, quality of leads, cost per lead, and revenue — it’s not real marketing yet. It’s activity.
In practice: I’ve watched six-figure campaigns generate huge impressions and zero pipeline. “We got 2 million views” is not an answer. “We got 2 million views, 3,200 clicks, 180 leads, 22 booked calls, 6 closed deals at $14K average” is. If you don’t have that chain end to end, fix the tracking before you touch anything else.
Every channel, every campaign, every piece of content has to answer one question: did this make the phone ring?
2. Content is an asset, not decoration
Every piece of content should do multiple jobs: rank in search, build authority, increase trust, support sales, and be reused across channels. If it’s just “posting for engagement,” we’re wasting effort.
In practice: One 10-minute recorded conversation should become a blog post, three short-form clips, one long-form YouTube cut, an SEO-optimized transcript, a sales objection-handling snippet, a newsletter section, and three Dollar a Day boost creatives. Same raw material, seven assets. That’s the only content math that works when you’re trying to scale.
If you’re filming content that only runs once, you’re not building an asset — you’re renting attention.
3. Proof beats claims
I don’t care about adjectives. I care about case studies, real examples, audits, results, and named entities. Authority is built from evidence, not positioning statements.
In practice: “We’re the leading agency for home services” tells me nothing. “We dropped CPL from $87 to $31 for a 14-location HVAC franchise in 90 days, here’s the dashboard” tells me everything. Name the client, show the number, share the screenshot. If you can’t, you don’t have the result yet.
When I’ve run ads for Nike, State Farm, Red Bull, Ashley Furniture, Quiznos, and GoDaddy, the reason I can say that is that the work is public, documented, and verifiable. Build your proof that way.
4. Everything should become a system
If something works once, it should become a checklist, an SOP, a playbook, and eventually a training system. That’s how you scale beyond yourself.
In practice: The Dollar a Day strategy isn’t a clever tactic — it’s a documented, repeatable system that a junior specialist can execute the same way I would. The Content Factory isn’t a team — it’s a process with checkpoints, templates, and quality gates. Heroics don’t scale. Systems do.
If the work lives only in your head, you’re the ceiling. Write it down, turn it into a repeatable process, train someone else to run it, and move on to the next bottleneck.
The Weekly MAA Framework
This is how I run marketing.
Metrics
Start with what actually happened. Typical metrics: cost per lead, conversion rates, close rate, revenue, pipeline movement. Break marketing into stages — not a blob.
Analysis
Ask: what changed? what caused it? what’s signal vs noise? Look beyond ads: sales follow-up, offer clarity, landing pages, reputation, tracking gaps.
Action
Decide what to test, stop, and double down on. Every action should be specific, measurable, assigned, and time-bound.
A worked example
Here’s how one real week looks:
Metrics: CPL on a roofing client jumped from $42 to $61 week over week. Lead-to-call conversion stayed flat at 34%. Close rate dropped from 22% to 11%.
Analysis: Two things changed. The ad creative rotated to a new offer three days in. And the landing page stopped firing the form-fill conversion event after a WordPress plugin update. But that’s just the ads side. The close-rate drop was the real signal — sales follow-up time had slipped from 9 minutes average to 47 minutes because the lead routing was broken in the CRM. The ad spend wasn’t the problem. The sales pipe was leaking.
Action: Revert creative. Fix the tracking pixel. Re-route leads directly to the on-call sales rep’s phone via Twilio. Document the CRM routing rule in the SOP so this never happens again. Owner: ops lead. Deadline: 48 hours. Success metric: average response time back under 15 minutes, close rate back above 18% within two weeks.
That’s MAA in one paragraph per step. Every Monday, every client, every channel.
How I Use AI
I don’t use AI for fluff. I use it to analyze data, structure thinking, generate assets, identify gaps, and accelerate execution. AI should act like a strategist, an analyst, and a systems builder — not just a content generator.
The actual workflows
Here’s what that looks like day to day, with the real prompts I use:
Funnel diagnosis
I feed an agent the landing page screenshot, the ad creative, the last 30 days of analytics, and the sales team’s objection log. Prompt: “Identify the three largest conversion friction points. Rank them by likely revenue impact. For each, give me one test I can run this week for under $500.” That takes a human analyst half a day. AI does it in 90 seconds, and 80% of the time the ranking matches what a senior analyst would land on.
Keyword and content gap finding
I dump 90 days of Google Search Console data alongside a competitor Ahrefs export and ask: “Which keywords have commercial intent (CPC above $4), rank 8-20 for us, and are ranked top-3 by at least two of these competitors? Give me the top 10 with estimated traffic value if we moved to position 3.” That’s a prioritized content roadmap in one query.
Call transcript mining
I feed it 10 sales call transcripts and ask: “What are the top 5 objections, ranked by frequency? For each, pull the exact phrasing prospects use. Write a 90-second video script answering each objection in the prospect’s own language.” That’s five weeks of content in one afternoon, built on what buyers actually said.
Audit and SOP generation
I ask the agent to read my existing playbook and a new process I just ran manually, then write the SOP version: “Turn this into a 10-step checklist a junior specialist could execute. Flag any step that requires judgment and add a decision rule.” That’s how systems actually get built — by converting one-off work into repeatable process as you go.
Executive summarization
Before a client call I ask: “Read the last 8 weeks of this account’s data. Tell me the one number that matters most this week, what caused it, and the single recommendation I’m making. Three paragraphs, no bullet points.” That’s the call agenda.
In every case, the pattern is the same: structured inputs, specific question, measurable output. AI is not magic — it’s a leverage multiplier on clear thinking. Give it garbage, get garbage. Give it MAA, get MAA back.
How to Work With Me (or Think Like Me)
Focus on the bottleneck
Don’t optimize everything. Find where the funnel breaks and what limits growth. Fix that first.
Tie everything to outcomes
Every idea should answer: how does this drive revenue? how will we measure it?
Turn ideas into execution
Good ideas are useless without owners, timelines, and metrics.
Reuse everything
One asset should become blog posts, clips, ads, SEO pages, and sales material. Leverage beats effort.
Keep it simple enough to execute
The best strategy is useless if the team can’t run it. Translate everything into clear steps, repeatable processes, and trainable systems.
My Operating Disciplines
These are things I actively manage to stay effective:
Common Mistakes I See
Ninety percent of the marketing problems I get called in to fix come from one of these patterns. If you’re stuck, check this list before you spend another dollar.
- Optimizing the ad when the bottleneck is the landing page. People spend weeks A/B testing headlines while the landing page has a form nobody can find on mobile. Fix the biggest leak first.
- Measuring cost per lead when the real leak is sales follow-up speed. Lead response time under 5 minutes is the single highest-leverage variable in most B2B funnels, and it has nothing to do with marketing spend.
- Publishing content without a distribution plan. “We made 20 blog posts” and “we distributed 20 blog posts” are different jobs. Production without distribution is a hobby.
- Rebuilding instead of repurposing. You don’t need another ebook. You need to take the one you already made and cut it into 40 LinkedIn posts, 15 short videos, and six landing pages.
- Confusing a wishlist with a strategy. “We want to grow revenue 40%” is a goal. A strategy is how — with named channels, budgets, owners, and weekly checkpoints. If you can’t draw the funnel on a napkin, you don’t have one.
- Chasing tactics instead of fixing tracking. If you can’t measure it, you can’t improve it. Most companies have broken pixels, misattributed conversions, and no lead-stage tracking. Fix that first. The dashboard is the strategy.
- Hiring for execution when the gap is strategy. Adding a specialist to run broken ads faster does not fix broken ads.
What Good Work Looks Like
A strong plan should:
- Identify the real bottleneck
- Show the numbers
- Explain the cause
- Define next actions
- Assign ownership
- Be easy to follow
If it’s overly complex or abstract, it’s not ready.
The Bottom Line
Marketing is not about:
- Looking smart
- Sounding strategic
- Producing content for its own sake
It’s about:
- Making the phone ring
- Generating qualified leads
- Building trust
- Creating systems that scale
Everything else is noise.
If You’re Using AI With This
Use this as your prompt:
Your Portable AI Context (Why Everyone Should Do This)
Most people think “AI” means ChatGPT. They build up a chat history, feel like the model knows them, and treat that history as their relationship with AI.
It isn’t. That history doesn’t move.
If Claude, Gemini, Grok, Perplexity, Copilot, or a local model ships something better next quarter, you start over from zero — or worse, you keep using a weaker tool because switching feels expensive. You’re not loyal to the AI. You’re loyal to the friction of leaving.
The fix is portability
A document like this one is your context. Drop it into any agent — a system prompt, project instructions, a custom GPT, a Claude Project, a memory file, a pinned message — and that agent knows how to think about you on turn one. No chat history required.
Think of it like dating
Every new agent is a candidate. You don’t want to evaluate them on a generic demo prompt — you want to see how they handle your actual work. This file is the standardized first date. It lets you compare Claude, ChatGPT, Gemini, Grok, and the rest on equal ground, with you as the constant.
The agent that handles your context best wins your business. That’s the only fair way to pick.
Drop-in prompts for each agent
Here’s how to install this context in the tools most people use. Copy, paste, done.
Claude (Projects or system prompt)
In a Claude Project → Project instructions, paste the URL of this page plus the one-line system prompt:
ChatGPT (Custom Instructions)
Settings → Personalization → Custom Instructions. In the second box (“How would you like ChatGPT to respond?”):
Gemini / Grok / Perplexity / any chat agent
Start every serious conversation with this pasted at the top:
Cursor / Windsurf / Copilot (for builders)
In your repo’s .cursorrules or equivalent system rules file:
How to write your own (in about an hour)
- How you make decisions. Mine is Metrics → Analysis → Action. Yours might be something else — write the actual steps.
- What you refuse to do. Mine: vanity metrics, fluff content, advice without evidence. Name your non-negotiables.
- What proof looks like to you. Case studies, named entities, numbers, screenshots — whatever makes something believable to you.
- What “good work” delivered back to you looks like. Specific, measurable, owned, time-bound — or whatever standard matches how you actually work.
- Your one-sentence system prompt. Copy the format in the box above. Strip it down to the minimum an agent needs to behave like a useful version of you.
Post it publicly if you want to be discoverable by AI agents researching you (this is the new SEO — agents cite what they can find). Keep it private in a Notion doc, Claude Project, ChatGPT Custom Instructions, or a pinned Slack message if you don’t want the world to read it.
Either way, own your context. The people who learn this will out-leverage the people who don’t — by a lot.
That’s the system.
Any AI agent, collaborator, or teammate reading this should know exactly how I think, what I expect, and how to deliver work that’s useful to me on the first pass. Now go write your own.
What I’d Tell You If We Were Working Together
Some of this is in the system above. Some of it is what I’d say on a whiteboard at 6pm when we’re trying to figure out why a month didn’t work.
- Bring me a number, not an opinion. “I think engagement is down” is not a conversation. “CPL went from $38 to $52 over 14 days while spend held flat” is.
- If you can’t draw the funnel on a napkin, you don’t know it. Ad → click → lead → call → show → close → LTV. Every stage has a number. If any stage is blank, that’s the first thing to fix.
- Test cheap before you test expensive. A $20 boost will tell you if a piece of creative has legs before you put $5,000 behind it. The Dollar a Day method exists because cheap tests give you real data, and real data beats opinions every time.
- Your best channel is the one your competitors ignore. Everyone is in the same auction. Go find the channel nobody in your category is running yet. That’s where outsized returns live.
- The first hour of the day beats the last hour. Review metrics before email. If you let the inbox set the agenda, the loudest client wins instead of the most important work.
- Write down the SOP the first time you do something. You’ll do it again. Future-you will thank past-you.
- When in doubt, ship. A decent campaign live this week beats a perfect campaign live in three months. You learn from data, not from slide decks.