The Senior Marketing Executive is becoming a Senior Marketing Engineer.
For twenty-five years, I led marketing with judgment, relationships, and budget. That still matters. But the leaders who will matter in the next twenty-five are the ones who can also read marketing as a system — and diagnose which layer is actually broken.
The expensive mistake isn't the wrong tool. It's working on the wrong layer.
Every week I watch marketing leaders swap ChatGPT for Claude, rewrite a prompt for the fifth time, or replace an agency because the output "feels off." Sometimes it helps. Usually it doesn't. And they can't explain why.
Here is why: marketing is no longer a single discipline you manage. It is a system with layers. The symptom lives on the surface — bad copy, a campaign that won't convert, an agent that forgets the brief. The cause usually lives three layers down.
Changing the tool when the problem is in the data layer is like rewriting a brief when the CRM is broken. You will lose the quarter.
Eight layers. Not one discipline.
Every modern marketing system is built from eight layers stacked on top of each other. You don't need to master all eight. You need to know which layer you're working on — and which three are where the money actually is.
The atoms. What it costs, how much the AI can see at once, how predictable the output is. Skip this layer and you pay for surprises.
Your AI is only as good as what it knows about your business. If it doesn't know your ICP, your pricing, your objection library — it will sound like every other vendor on earth.
The rules the AI operates under. Brand voice, what it will and won't say, what it remembers between conversations. This is where "sounds like my company" actually happens.
The engine. Different models for different jobs — not a single winner. Most of your competitors are stuck debating this layer. It's the least valuable place to spend time.
How your systems talk to each other. When your CRM, email platform, data warehouse, and AI all work in concert, you get compounding output. When they don't, you get a bill.
Where a tool stops being a tool and becomes a worker. This is the layer that replaces entire job functions — not people, but the busywork that used to require them. This is where real businesses are being built.
The toolbox that lets a non-engineer ship production systems. The right tool for the right job is the difference between a three-day build and a three-hour build.
The lens. If a build doesn't save time, add revenue, or cut cost — it shouldn't exist. Every layer above has to answer to this one, or you've built a hobby.
A token is a budget line item.
Every word you send to an AI and every word it sends back is counted. Roughly one token per three-quarters of a word. For a marketing leader, this is not a technical detail — it is a unit economics question.
Running a one-off draft? Pennies. Running the same prompt across five thousand accounts in an ABM sequence? Now it's a budget conversation. The leaders who understand this layer build systems that scale. The ones who don't get surprise invoices.
The same principle applies to context windows — how much your AI can hold in its head at once — and temperature, which dictates how predictable its output is. Sales enablement needs low temperature. Brainstorming needs high. Both need to fit inside the context budget.
Generic in, generic out.
When a marketing leader says “the AI output is bland,” they almost always assume a prompt problem or a model problem. It's rarely either. It's a data problem.
An AI with no access to your CRM, your voice of customer, your pricing logic, or your closed-won history is going to produce the average of the internet. Which is exactly what it sounds like. If your output feels like a competitor could have written it, the layer to fix is this one.
The engineering move is a knowledge layer — your company's institutional knowledge, structured and made searchable by the AI. Embeddings, vector databases, retrieval. The technical terms don't matter. What matters is that your AI writes like it works for you instead of for nobody.
- The public internet
- Generic industry framing
- Your competitors' websites
- Your ICP
- Your pricing structure
- Your objection library
- Your win/loss patterns
- Why customers actually buy
- All of the above
- Your CRM history
- Closed-won call recordings
- Segmented persona libraries
- Battle cards by vertical
- Pricing & discount logic
- Voice of customer corpus
- Real competitive intel
The gap between these two columns is the gap between “generic AI output” and “this sounds like it came from inside our company.”
The layer that gives your AI a job description.
A model without an intelligence layer is a very capable intern with no onboarding. It will do the thing you asked for and also thirty things you didn't. The intelligence layer is where you tell it what it is, what it does, and what it will never do.
Three pieces do the work. The system prompt defines the role — tone of voice, what the brand stands for, which frameworks to use. Memory lets it accumulate context across conversations the way a senior employee would. Guardrails set the hard rules: never discount below X, never promise features we don't ship, never answer legal questions.
This is also the layer where most customer-facing AI disasters happen. An AI assistant without guardrails is a brand liability. An AI assistant with them is an employee.
Don't pick a model. Pick a job.
This is the layer your competitors waste the most time on. The debate over “which AI is smartest” is a conversation for people who haven't built anything yet.
There is no best model. There are right models for specific jobs. One model writes strategic narrative well. Another runs orchestrated agents better. A third lives inside your Google stack. A fourth runs on your own hardware.
The senior marketing engineer matches the model to the work. The senior marketing executive who hasn't made the shift is still arguing about ChatGPT versus Claude at dinner.
Long-form strategic writing, nuanced sales narratives, editorial judgment. Strongest when the output needs to read like a senior operator wrote it.
Versatile general-purpose workhorse. Strong at structured data extraction, tool-calling, and agent orchestration.
Deep integration with Google's ecosystem. Useful when the workflow is already inside Workspace, Ads, or Analytics.
Open source. Full control over where the data lives. Strong choice when compliance, privacy, or custom fine-tuning matters more than raw capability.
Lean, efficient, the European alternative. Cost-competitive for high-volume tasks where marginal quality loss is acceptable.
The compounding layer.
An AI in a browser tab saves an hour. An AI connected to your CRM, your email platform, your data warehouse, and your calendar saves the afternoon, every afternoon. The difference is connectivity.
APIs let systems ask each other questions. Webhooks let systems notify each other when things happen. MCP — the Model Context Protocol — is the newest and most important development in this layer: a standard way for AI models to plug directly into the tools they need.
When a marketing leader says “we bought a lot of tools but nothing actually compounds,” they are describing a connectivity problem. The tools work. They just don't talk. This is the layer where the most hidden ROI lives in most organizations.
This is where the game changes.
A prompt does one thing. An agent takes instructions, breaks a job into steps, uses tools, and comes back with a result. A multi-agent system runs several specialized agents together — one researches, one writes, one reviews, one dispatches, one logs to the CRM.
This layer is not a marginal improvement. It replaces entire process surfaces. The work that used to take an SDR two hours per account now takes five minutes of human review. The team stays the same size. The output multiplies by ten.
Most marketing leaders know this layer exists and still haven't deployed it. That gap is the single largest source of competitive advantage in marketing right now.
- 45 minResearch account
- 20 minPull firmographic data
- 15 minCheck LinkedIn, news, triggers
- 30 minDraft personalized sequence
- 15 minReview against battle card
- 10 minLoad into CRM
- 10 minSchedule send cadence
- autoResearch agent pulls firmographics
- autoIntent agent scans news + signals
- autoCopy agent drafts sequence
- autoQA agent runs voice + brand check
- 5 minHuman reviews & approves
- autoCRM + dispatch agent writes & sends
- autoAttribution agent logs to pipeline
The leader doesn't need to code. The leader needs to know which tool to pick.
Ten years ago, deploying a custom marketing workflow meant hiring developers. Today a marketing leader can ship production systems with a handful of no-code and low-code tools — provided they know which tool to reach for.
The toolbox is not infinite. Six or seven tools cover most of what a modern marketing team needs to build. The difference between a three-day build and a three-hour build is almost always the same: the right tool, selected by someone who knows the landscape.
This is what “senior marketing engineer” actually means in day-to-day terms. Not writing code. Making architecture decisions about which systems compose the stack.
Self-hosted, open-source automation. Full control, no per-task pricing, strong AI-native support.
Visual multi-step workflow builder. Sweet spot for complex logic without writing code.
The gateway drug. 6,000+ app integrations, low learning curve, first automation win in a day.
AI-powered code editor. Describe what you want, get working code. Non-developers ship real systems here.
No-code builder for conversational AI and customer-facing agents. Drag, drop, deploy.
Open-source visual builder for RAG pipelines, custom agents, and LLM apps. Self-hostable.
The lens every other layer has to pass through.
A marketing engineer who can't tie every build back to revenue, time saved, or cost cut has not built a system. They have built a hobby.
The business layer is where the stack meets the board. It asks three questions of every build on every layer above: What is the use case? What is the ROI? Does this belong in our AI strategy, or is it random tool adoption?
Start with the problem, not the product. Every engagement I've run in the last two years starts here — and works upward through the stack only after this layer is settled. That is the difference between a marketing team with AI bolted on and a marketing operation that was engineered for this era.
Senior Marketing Executive. Senior Marketing Engineer. The same role, twenty-five years apart.
Both still require judgment. Both still lead teams. Both still answer to revenue. The difference is what's beneath the strategy deck.
“I spent twenty-five years learning to lead marketing through judgment, relationships, and revenue instinct. I spent the last two learning to engineer it. The combination is the craft.”
The senior marketing leaders I work with are not trying to become developers. They're trying to stay relevant in a moment when the craft of marketing has split into eight layers instead of one — and the companies that win will be led by people who can see all eight.
This page is the shorthand for how I think about every engagement. When a CEO brings me in, my first job is diagnostic: which layer is the pain actually on? The second job is to build the system, not the strategy deck. The third is to leave the operation better engineered than I found it.