The Marketing Engineering Era

Picture an independent optometry practice.

Right now the marketing works like this: post when somebody remembers, boost whatever felt good that morning, then wonder in January why the year came in flat.

Nothing about that is laziness. Every review, every question at the front desk, every phone enquiry that practice takes is market research. It arrives all week long and it evaporates by Friday, because nothing is catching it.

Now picture the same practice with one thing added. Everything patients say lands in one place. Once a week, something reads all of it and tells the owner what changed, with the actual quotes attached. And what it reports is that almost nobody is asking about frames. They're asking whether you take their insurance and whether the appointment will run over an hour.

That single sentence rewrites the homepage, the next six posts, and the first line of the voicemail. Then the system watches what those changes do, and reports again next week.

Same owner. Same budget. The practice can finally hear its own market.

That gap is what this post is about, and there is now a name for the person who closes it.

Four eras, and we just entered the fourth

Greg Isenberg published an episode this week making the case that the next genuinely valuable role in marketing is the marketing engineer. Some people are calling it a forward deployed marketer, some an AI growth operator. His argument is that the label will keep moving but the work will not.

He lays the shift out in four eras, and the sequence is his, not mine.

Traditional marketing was about making people care. Print, radio, television, and a copywriter who understood what you wanted, what you were insecure about, and who you were trying to become.

Digital marketing arrived with websites, email, search and social ads. The winning marketer became whoever could acquire customers through channels you could finally measure.

Growth hacking pulled marketing into the product. Activation, referral, onboarding, retention, pricing. The product itself became the engine.

Marketing engineering is the one we just walked into. It keeps all of the above, because you still need judgment, positioning and taste. What's new is that this marketer also builds the system underneath the marketing. His definition is the cleanest I've seen: the person who turns market signal into pipeline using AI agents, data, code and taste.

The word carrying the weight there is system. A campaign ends and a content calendar just repeats. This is something that gets smarter every week whether or not anybody remembers to make it smarter.

The difference between using AI and building with it

Here's the part most business owners are missing, and I say that with sympathy because it took me a while too.

Most people use AI in scattered conversations. You open a chat, ask for ten captions, keep the two you like, close the tab. Next week you start from nothing. The model has no idea which of last month's posts brought in an actual client, what your best customer said on the phone on Tuesday, or which three phrases you'd never let your brand say out loud.

The work evaporates every single time.

A marketing engineer fixes that by giving the system a memory. Isenberg's starting point is what he calls a growth repo, a structured folder where the company's marketing memory lives. Customer truth in one place. Voice and winning hooks in another. Ideal client profile, and the language you've banned. Creative tests and what they returned. A file for each agent describing its job.

Then the instruction stops being "write me ten posts" and becomes "read what our customers actually said this month, read the five posts that produced real inquiries, and draft five more about the pain they named this week."

Same tool. Completely different output. The difference is not the model. It's the memory.

What this looks like when it is actually running

I want to be concrete, because this subject attracts a lot of talk and very little proof.

My own studio runs as a house of specialists. Each one is a written job description rather than a personality: what it reads, when it runs, what it must never do, what good output looks like, what has to come to me for approval, and where it writes its results so the next run starts smarter than the last. A copy inspector reads every client facing piece before it ships and returns a numbered fix list. A separate inspector checks anything visual. A third checks live site edits. A running ledger tracks recurring mistakes, and any fix proposed three times without a decision stops being re-proposed and comes to me as a straight keep or kill.

Three of the inspectors. Nothing client-facing ships without passing at least one of them.

None of that is glamorous. All of it is the actual job.

And here's the honest part, which usually gets left out of posts like this one. This week that same system handed me a problem it had first logged thirty four days earlier. When I pulled on it, two of its three parts had never been broken. One had been working for some time already and nobody re-checked. One was a guess that hardened into a stated fact, because the test that would have disproved it could not run.

The ledger keeper. She is why a repeated mistake gets named and put in front of me instead of quietly repeating.

Building the system was still the right call. Building it with no way to re-check itself is what went wrong. A list that only ever gets added to will accumulate confident fiction, and deliver it on a schedule in a clean template where it reads exactly like fact. The automation is the easy half. The checking is the job.

Why this matters more for a boutique brand

Most of the conversation about marketing engineers is aimed at funded startups hiring a specialist at a very large salary. Isenberg thinks the best of them will earn well past a million a year, and I think he's right about the ceiling.

But the gain is bigger at the small end, and almost nobody is saying so.

A venture backed company already has a growth team, a data person and an agency. Adding a marketing engineer makes a capable team faster. A boutique practice or a founder led brand has none of that. The owner is the marketing department, between clients, at nine at night. For them this system decides whether marketing happens at all, or keeps getting postponed to a quieter week that never arrives.

That's the optometry practice at the top of this post. The same owner, on the same budget, hearing clearly for the first time.

Systems and taste, and why one without the other fails

There are two halves to this job and they attract very different people.

The systems half is the folders, the job specs, the scheduled runs, the approvals, the checking. Unglamorous, learnable, and where most of the durable advantage sits.

The taste half is knowing what should exist at all. Which pain is worth naming. Which sentence sounds like a real person and which sounds like software wearing a person's coat. When a piece is technically correct and still wrong.

The point Isenberg lands on is the one I keep returning to. His argument is that agents become a commodity, and what stays scarce is knowing where to point them.

That's exactly right, and it's why I'm not nervous about any of this. The tools get cheaper and better every month, which means average marketing is about to be nearly free and completely forgettable. What doesn't commoditize is knowing what deserves to be made at all.

I spent thirty-three years in luxury wholesale account management, sitting across the table from small business owners, watching what made one boutique feel considered and the one next door feel generic. That's where the taste came from. The systems came later, and I built them because I needed them.

Neither half works alone. Taste without a system produces a beautiful brand that posts four times a year. A system without taste produces a great deal of perfectly grammatical content that nobody remembers reading.

Where to start, if you want to

You don't need a repo, a team, or a single line of code to begin. You need one honest hour.

Open a document. Paste in the last twenty real things your customers said. Reviews, emails, the questions they ask before booking, the objection you hear so often you've stopped noticing it. Then ask: what changed this month, show me the evidence, and what's the one thing worth testing this week.

That's the first version of a customer truth file, and it's more than most brands have.

If it earns its keep, the next step is putting it on a schedule instead of running it when you remember. That's the whole path. A habit that compounds.

The window

Most businesses are still using AI the scattered way, one conversation at a time, starting over every week. That's why there's room to move right now, and why I would rather write this today than a year from now.

Everyone has the tools. The advantage is in building the memory, pointing it at the right thing, and having the taste to know when the output isn't good enough to carry your name.

That last part still can't be automated. I don't think it ever fully will be.