You Don't Have to Learn Claude Code to Run an AI-Native Marketing Org

I've had something like twenty conversations about the Cursor CMO series in the last two months, and a pattern has emerged that's worth naming out loud.

The conversations start the same way, every time. Someone reads the series (the anchor piece, the strategy chain from discovery to positioning, the ICP piece on synthetic customers) and they book time. They open the call already engaged. They want to talk about specific skills they liked. They ask detailed technical questions about the MCP server, the skill files, how the agents are orchestrated. They're genuinely impressed with what they see, and (honestly) a little wound up about it in the best way.

Then, about fifteen minutes in, we hit the thing.

It's never "is this legit?" It's never "does this actually work?" They're already sold on those. The real question, almost every time, is some variation of:

"Do I have to know as much as you do to get value from this?"

"Am I going to have to put 1,500 hours into this before it pays off?"

The worry isn't about the validity of the approach. It's about the personal entry cost. These are smart people who've looked at what I've built, done the math on the level of fluency I'm operating at, and landed on a rational conclusion: this is real, it's impressive, and I don't have the bandwidth to become that person.

The good news, which I want to say as clearly as I can: you don't have to.

There are multiple legitimate paths into AI-native growth, and the path I took is only one of them. It's not the right path for most people. It's barely even a path I'd recommend to most people. It just happens to be mine because I'm a specific kind of operator with a specific set of constraints (a founder, technical, obsessed, the kind of person who enjoys debugging a Claude Code script at 11pm on a Tuesday because it's genuinely fun). Most of you are not me. Most of you shouldn't be.

What you should do instead is pick one of three legitimate paths and commit to it. That's the whole piece.

In this piece:

• The pattern I keep hearing, and what it actually means about personal entry cost
• The Three-Lane Rule™, which is the three service modes for engaging with AI-native growth (spoiler: one of them is "pay someone else")
• The Claude Code Cosplay trap, which is what happens when you assume the creator's path has to be your path
• The fluency gradient, which is my way of saying you don't have to be at the bottom of anything
• How to pick your lane in four honest questions
• Why my business is structured the way it is, and which door is actually yours

You can drive a car without knowing how to build one

Here's the line I keep coming back to in these conversations. You can drive a car without knowing how to build one.

Telling every marketer they now have to become a growth engineer is like telling every driver they have to be a mechanic. You can drive across the country without knowing how a fuel injector works. You can change a flat without rebuilding the transmission. You can tell when something's wrong and hand the car off to someone who knows more than you. None of that requires popping the hood every morning (or, frankly, ever).

Most people who work in marketing are going to be drivers. A smaller group will be home mechanics (the ones who like popping the hood on weekends, who know what a fuel injector is, who voluntarily subscribe to forums). A much smaller group will be professional mechanics who actually build and maintain the vehicles the rest of us drive. All three of these roles exist inside a functional ecosystem. None of them is "better" than the others. They do different jobs.

There are two versions of this trap, and they're both active.

The external one is loud. Tech Twitter, the "build-it-yourself-or-die" hype machine, the course sellers, the "X is dead" clickbait merchants. That version flattens everything into one message: if you aren't rebuilding the engine, you aren't participating. I'm not the first person to call this out, and I won't be the last.

The internal one is quieter, and I think it's the more damaging of the two. It's what happens when you look at someone doing genuinely interesting technical work, correctly identify that the work is valuable, and then assume the only way to capture that value is to become them. You don't ask whether a shorter path exists. You just start the long one, get overwhelmed somewhere around month two, and quit by month four.

I'm calling this the Claude Code Cosplay trap. It's performing growth engineering because you watched someone else do it, without pausing to ask whether the role actually fits your body, your team, and your organization. Dressing up as the thing, instead of asking what the thing is actually for.

Both versions share a root cause: the assumption that there's one right way to go AI-native, and it's whichever way the loud people on the internet (or the creator you happen to read, hi) happen to be doing it. There isn't one right way. There are three.

So the real question isn't "Do I need to learn Claude Code?"

It's "Which of the three paths actually fits my situation?"

The Three-Lane Rule™

There are three service modes for engaging with AI-native growth. Pick one. They look very different, cost very different amounts, and fit very different businesses.

(Yes, I'm naming it. If Emily Kramer gets a LinkedIn Flywheel™, I get a Three-Lane Rule™. We don't make the rules, we just trademark the frameworks.)

Lane 1: Do It Yourself (DIY)

You, or someone you hire into your marketing org, builds the system. You own it. You operate it. You iterate on it weekly.

This is the lane I get asked about the most, almost always by people who are in the wrong lane for it. The pattern looks something like: VP Marketing at a 400-person mid-market company watches a Cursor CMO demo, gets genuinely excited, goes back to the office and tries to teach a 12-person marketing team how to use Claude Code. Six weeks later they're burned out, their team is resentful, and nothing shipped.

Who this actually fits:

• Startup marketing teams, often under five people total
• Orgs without a meaningful MarTech budget (nobody's about to greenlight $3K-a-month platform licenses)
• Product and engineering teams that actively support marketing shipping its own vibe-coded tools (this part is non-negotiable)
• Teams in iteration mode: pre-PMF, pivoting strategy monthly, allergic to a six-month system build
• Leaders who are technical themselves, or who can hire growth engineers directly onto the marketing team

What you do Monday: Hire someone who can ship in Claude Code, or start learning yourself if you're the founder. Build the three or four skills you use most (content production, social scheduling, ads analysis, whatever your highest-frequency work is). You're not buying the $3K-a-month SaaS. You're building thin, fit-for-purpose versions of it and improving them weekly.

What it costs: Time. Your time, or your first growth hire's time. Budget 30%+ of their week on system development for the first six months. After that, the system compounds and they get most of their time back for strategy and output.

What you give up: Polish. None of what you build will have a nice GUI. It lives in terminals and folder structures and Git repos. If your team is non-technical, they can't touch it. That's a real tradeoff, and for the right team, it's worth every pixel they didn't get.

Lane 2: Done With You (DWY)

You bring in an operator or studio to build the system for you, tune it with your team, and teach you how to run it. The system becomes yours. A good DWY partner will also help you find and onboard the growth engineers who take over after they leave.

This is the lane most mid-market leaders actually belong in, and it's also the one almost nobody names out loud. The conversation usually defaults to a binary (either build it or buy an agency), which skips right over the most useful option.

Who this fits:

• Middle-market businesses, often PE-backed, somewhere in the $50M-$600M revenue range
• Teams with real MarTech budget (six or seven figures annually) but no established growth engineering practice yet
• Orgs with a predominantly non-technical marketing team, where asking everyone to learn Claude Code would cost you 30-40% of team productivity for half a year (a real number I have watched companies spend)
• Leaders who know they need to modernize but don't want the first build attempt to be the thing that proves it's possible

What you do Monday: Don't try to teach your whole team Claude Code. Start by upgrading pieces of your MarTech stack to AI-native tools. That's your training ground. Your team gets comfortable with LLMs inside software they already understand. Get Claude Cowork licenses across the team and start a Slack channel where people share the skill files that save them time. Hire one or two technical folks (or bring in a studio like mine) to seed a growth engineering practice internally. Over the next 6-12 months, you build a hybrid team: most people working in AI-upgraded SaaS, a small crew building custom solutions where off-the-shelf doesn't fit.

What it costs: Real money up front. A DWY engagement isn't cheap. It is dramatically cheaper than a failed internal attempt, and the asset that stays with you (the system plus the people who know how to run it) is worth more than the fees.

What you give up: Full ownership during the build. For the first 6-12 months you're working inside someone else's mental model. Most mid-market leaders find this is a feature, not a bug.

Lane 3: Done For You (DFY)

You hire a growth agency that runs on this kind of system under the hood. You never see the inside. You never own it. You buy the output: campaigns, content, ads, reports.

I'm going to say something mildly heretical here, which is that this lane is fine. The growth engineering world likes to look down on the agency path, but for a lot of businesses, this is the most rational move. Marketing ops doesn't have to be your competitive advantage. It's okay to outsource it.

Who this fits:

• Teams whose comparative advantage is somewhere other than marketing operations (this is most teams, honestly)
• Businesses in scale mode, with a proven playbook, where execution quality and volume matter more than strategic iteration
• Leaders who want AI-native results without adding any internal complexity
• Orgs where the exec team explicitly does not want marketing to own internal systems

What you do Monday: Interview agencies the way you'd interview any other vendor, with one non-negotiable requirement. Ask them to show you the system. If they can't, they're selling AI theater. Ask what happens if you leave. Ask what stays with you. Look for agencies where the system is the product, not the billable hours.

What it costs: Ongoing fees, which tend to run higher than traditional agency pricing at first because you're paying for the system, not just labor. Per-piece economics usually improve dramatically once volume kicks in.

What you give up: You never learn the system. That's the trade. You're buying freedom from operational complexity, and the price is some dependency on the vendor who has it. Negotiate for portability up front and this becomes a much smaller concern.

The fluency gradient, or: what kind of driver are you?

Even inside each of these lanes, there's a range of how deep you personally need to go. Think of it like driving levels.

Level 1. Daily driver. You use ChatGPT in a browser. You use the AI features that shipped inside your existing SaaS. You're AI-native in your daily workflow. This is where most marketers are, and it's a legitimate place to stay for your entire career.

Level 2. Knows the dashboard. Claude Cowork is open during your workday. You've built a handful of skill files for things you do repeatedly. You share them with teammates. You evaluate AI-native MarTech options when it's time to re-up a contract.

Level 3. Home mechanic. You're writing custom skills. You've set up an MCP server or two. You're building small agents to handle specific workflows. You can read Claude Code output even if you can't fully write it from scratch yet.

Level 4. Professional mechanic. You ship production software. You build agents, orchestration systems, internal tools that replace whole categories of SaaS. This is growth engineering, full stop.

Your lane determines how deep you personally need to go.

• DFY (Lane 3): Level 1-2 is all you need. Your agency handles Levels 3 and 4 on your behalf.
• DWY (Lane 2): Level 2-3 is the target. You're running a system someone else built, which means you need to understand it, not construct it from raw parts.
• DIY (Lane 1): Level 3-4. You or your first hire is at Level 4. Everyone else on your team can happily live at Level 2.

Almost nobody needs to be at Level 4. The hype machine would have you believe everyone does. That is the lie at the center of the current narrative, and it's the reason so many otherwise excellent marketing leaders are walking around feeling inadequate for no reason at all.

How to pick your lane

Four questions. Answer them how things actually are, not how you wish they were.

1. How big is your team and what's your MarTech budget?

Under five people, minimal budget → DIY leans favorable. Five to fifty people, mid-six to low-seven-figure budget → DWY or DFY. Over fifty people, seven-figure budget → usually DWY for the system build, DFY for execution depth in channels where you don't want the complexity (paid media is a classic example).

2. Does your org actually support marketing shipping its own vibe-coded solutions?

This is the question that kills more DIY attempts than any other, and it's almost always answered aspirationally on the first pass. If your engineering team rolls its eyes every time marketing proposes building something, or if every small internal tool has to route through a six-month product review, DIY will not work in your org. You'll build the system, and the org will reject it like a bad transplant. DWY or DFY is the move.

3. How technical is your existing team, really?

Not "are they smart people." I know they are. The question is: are they comfortable with schema changes, CMS backends, tag management, Liquid scripting inside email platforms, building their own customer segments in the CDP? If yes, DIY is viable. If no, DIY will cost you six months of team productivity and probably fail anyway. DWY is specifically designed for exactly this team composition (it's the most common one I see).

4. Are you in iteration mode or scale mode?

Iteration mode (pre-PMF, strategy pivoting monthly, hunting for the right motion in a new channel) rewards DIY. You can't afford the build time of DWY, and DFY agencies are too slow to pivot alongside you. Scale mode (playbook running, quality and volume matter more than strategic experiments) rewards DWY or DFY.

[VISUAL: Decision matrix with 4 questions as rows, DIY/DWY/DFY as columns, showing which answers lean which direction]

If the answers pull you toward one mode, that's your lane. If they pull you in multiple directions, it probably means you have different needs in different parts of your marketing function. That's normal. Big companies routinely run DFY on paid media, DWY on content ops, and DIY on growth experiments simultaneously. The Three-Lane Rule™ isn't "pick exactly one forever." It's "be honest about which lane each piece of the function belongs in."

Why I built my business the way I did

Quick aside, since it's relevant. I've spent roughly 1,500 hours building this system in Claude Code over the last eighteen months (yes, I counted; yes, I'm a little embarrassed about it). I know exactly how deep the rabbit hole goes. I also know most marketing leaders should not be the ones going into it, and I structured my business to match the framework I just walked you through.

Marketer in the Loop is for Level 1-2 drivers who want to stay drivers. The newsletter. The MCP server with the skills already built, so you can install and use what I use without building it yourself first. The tutorials. If you want to DIY but don't want to start from zero, Marketer in the Loop skips you past the first 300 hours.

MultiplAI Growth is for DWY and DFY. If you want us to come in and build the system alongside your team, run it with you until you're comfortable, and help you find and onboard the growth engineers who'll own it after we leave, that's DWY. If you want us to run the motion for you while you focus on the rest of your business, that's DFY.

The reason the business is shaped this way is because the framework in this piece is real. There's no single right answer to "how should I go AI-native?" There are three right answers, and they map to three very different kinds of organizations. I built for all three because pretending there's only one was making smart people feel stupid. Enough of that.

What to do Monday

Stop feeling guilty about not being at Level 4. Most of you shouldn't be, and the people who are will cheerfully tell you they didn't choose it for prestige.

Pick your lane using the four questions. Be honest about the one that pulled you hardest.

If it's DIY, go hire someone technical (or become that person) and start building. Budget six months of foundation work before compounding kicks in. Don't rage-quit at month four (that's when everyone wants to).

If it's DWY, start with a MarTech audit. Where can you swap in AI-native tools as a training ground for your team? Which one or two technical hires do you need to seed a growth engineering practice? Who can run the build alongside you while your team learns?

If it's DFY, start interviewing agencies with that one non-negotiable requirement. Show me the system. Anyone who can't is selling you hours, not leverage.

In all three cases, do this today: get Claude Cowork licenses for your team. Start a Slack channel for skill files. Let your people become drivers. That part is free, it builds fluency across your org quietly over time, and it's the foundation under all three lanes.

You don't have to learn Claude Code to run an AI-native marketing org.

You just have to pick your lane.