Before you hire a marketing engineer

Marketing has a new job title. A marketing engineer is a marketer who turns recurring work into systems: the workflows, the agents and the context they run on, plus the judgement about where a person still has to sign off. Companies have started hiring for it, and the job ads are long.

My advice to most enterprises is to hold that job ad for now. Build marketing engineering as a team capability first, with one named owner for the shared context, workflows and quality gates, and train every marketer to build on that system. Then hire, because by then there's a system for the hire to own.

A new title for work that already existed

I started as a writer. For most of my career, the job of turning a team's know-how into something repeatable sat with whoever cared most about it: an editor with a style guide, an ops lead with a checklist or a content manager with a template folder nobody else opened. AI has made that work impossible to skip.

A model only has the standards someone has written down, and where nothing is written down it fills the gap with the average of everything it has read. So the person who writes the standards down, connects them to the tools and checks what comes out has become one of the most important people in a marketing team. Now that person has a title. (It's a different job from the technical marketing engineer you'll find at hardware and software companies, who explains complex products to buyers.)

The scepticism is fair. The Wall Street Journal has reported that some critics see the title as "a surface-level rebrand" of existing roles, and marketing has renamed jobs before. Whether or not the title lasts, the work behind it is already sitting on the desk of every marketing leader who has bought AI licences and is still waiting for the results.

What marketing engineer job ads ask for

Read a marketing engineer job ad closely and you find several jobs in one. Figma's listing asks for someone to build workflows and agents, to "coach and partner with marketers to take them from novice to full workflow transformation and self-sufficiency", and to own a "marketing AI cortex of prompts, skills, context files, and agents" that's run like a product. It even asks the hire to "document every workflow, tool, playbook, and transformation pattern so knowledge scales beyond any one person", and it offers a base salary of $127,000 to $296,000 for all of it. Enterprise listings can stretch further. Equiniti/Notified's Marketing Engineer & Director of AEO adds coding, technical SEO, production agents and managing a team of web managers and developers to the same role.

Between the two listings, that's a builder, a trainer, a systems owner and a manager in one person. They're good ads, because that is what the job involves.

The trouble starts when an enterprise tries to hire all of that into a team where nothing is written down yet. The new marketing engineer spends the first months interviewing colleagues about what good looks like, working out the brand voice from old campaigns and finding out which tools IT will allow. Whatever they learn lives in their head. When they leave, and good people do leave, the system goes with them.

I hold my own work to one test: if it only works when I'm in the room, it isn't finished. A marketing engineering function that depends on a single hire fails the same test.

Content engineering: what every agent runs on

Look under any marketing workflow or agent and you find the same few things. A clear account of who you're talking to and what you're selling. Voice rules a model can follow. Examples of good work and, just as useful, examples of bad work. A written specification of each step. Quality gates that say who checks what before it ships.

Content engineering is the name for that layer: the context, standards and checks every workflow and agent relies on. I think of it as the foundation of marketing engineering. Marketing engineering is the wider function, covering campaigns, data, ads, reporting and anything else a team might automate, and content engineering is the part all of it draws from, because every agent you deploy, whether it writes product copy, answers customers or drafts a campaign brief, is only as good as the context and standards it's given.

Content engineering is also the hardest part to hire for. Connecting tools is a skill a marketer can learn in weeks, while knowing what good looks like, and writing it down precisely enough that a model can be held to it, takes an editor's judgement as much as a builder's. I've argued before that AI forces the instincts of good communicators to become explicit standards. That writing-down is most of the job.

Teams come to me with what they think is a tooling problem and leave with a definition problem.

Start with what IT already allows

Most of the marketing engineering examples doing the rounds come from AI-native companies, where a team can try a new tool on Monday and have it connected to the CRM by Friday. Enterprise marketing works differently. Tools go through approval, some connectors are off the table, and nothing touches customer data until security has signed it off.

IT's limits tell you where to start. I've built content systems with no external model access, no connectors, an approved Microsoft stack and nothing else, and every time the constraint decided the order of the build.

The foundation is the part you can build today. Positioning, audience, voice rules and examples are documents. A written workflow is a document. A quality gate is a named person and a checklist. All of it works with whatever model your company already licenses, and all of it survives the next change of tool. Agents come last, once the single workflows run reliably.

Quality gates belong in the foundation from the start. A human stays in the loop wherever AI produces something, and once the volume is too big to read everything, review has to be sampled. I've built a sampling QA system for exactly that: a planned sample gets checked, every failure is logged with its cause, and the step that caused it gets fixed, so each round improves the system as well as the batch.

Sampled quality gates are what make volume possible. I've automated product descriptions for an enterprise catalogue where thousands of SKUs had none, a backlog eight people had been permanently behind on. For Graia's SEO content engine, I estimate a writer needed five to eight hours per article from research to publishing, against 15 to 20 minutes with the engine.

One owner, every marketer building

Once the foundation exists, someone has to own it. That person keeps the context current, decides which outputs need a human signature and looks after the quality gates everyone else inherits. I've written about the point where automating a process turns into staffing a position, and this is the person who staffs and supervises that workforce. In many enterprises they already work there: a content lead, an editor, a marketing ops lead with an eye for words. The title is optional.

Everyone else builds on top. Holly Chen, who leads growth marketing at Samsara, told Nick Lafferty, who holds the title himself as a founding marketing engineer at an AI search company, that "every single marketer needs to be a marketing engineer". Lafferty describes her team as a 230-person marketing organisation running roughly 90 agents, and adds that "someone owns the infrastructure those 90 agents run on". Both halves matter: every marketer building, and one person owning what they all build on.

Getting a whole team there is slow on purpose. I take one task at a time and move it one step up a ladder: an assistant that does a task you review, then an automation that chains several steps, then a workflow that reports how it's doing, and only then an agent that works towards a goal without hand-holding. Most teams should stay on the first rung longer than they expect. A well-designed assistant pays for itself.

So, should you hire a marketing engineer?

Hire a marketing engineer eventually, or appoint one from inside, once there's something to hire them into.

If your shared context, workflows and quality gates are written down and running, a marketing engineer has something to own and extend, and the job ad gets a lot shorter. If they aren't, the hire becomes the system and you're back to depending on one person.

Write the system before you write the job ad.

Both paths end with one owner. The difference is whether there’s a system for them to own.

Whichever way you go, the specification comes first. Writing it is what a Content System Blueprint does, and it ends in one of three honest outcomes: you build it yourself, you hire for it and the specification becomes the job brief, or I build it.

The long job ads are a good sign. Companies have worked out that AI in marketing needs someone who builds. Give that person a foundation to stand on and they'll spend their first months extending it.