My name is Kristina Vislianska. I'm a growth marketer with 10+ years in B2B and B2C SaaS, AI, and Web3, and for the last couple of years, most of what I ship isn't campaigns. It's systems that run campaigns.

At some point, recruiters and founders started calling this "marketing engineering." I had to Google it. Turns out I'd been doing it for a while; I just called it "building stuff so I don't lose my mind doing the same thing every Monday."

This guide is for marketers who want to build, founders who want to hire someone who builds, and engineers who wonder what marketers are doing with their Claude or Codex Pro Max. Everything here comes from systems I built and ran, including the ones that broke.

Key points, if you only have a minute:

• Marketing engineering turns repeatable go-to-market work into software, so one person can run what used to take a team.
• The work follows a five-layer loop: Listen, Understand, Make, Ship, Learn.
• Decide what to automate with two questions: how often does the task repeat, and how bad is a public mistake?
• The biggest wins are in listening and learning, not in generating more content.
• You don't need a CS degree. You need to read code well enough to know why it broke.

What is marketing engineering?

Marketing engineering turns repeatable go-to-market work into software, combining marketing judgment with engineering skills. Its systems listen to the market, turn what they hear into content, distribute it, and learn from the results. A marketing engineer's output is not a campaign. It's the machine that produces campaigns.

The deliverable is a monitor, a pipeline, an agent or a feedback loop. Something that keeps working when you close your laptop.

And no, it's not "marketing that uses AI tools." Asking ChatGPT for ten hooks is using a tool. Marketing engineering starts when the work keeps running without you typing.

The title keeps changing. A few years ago it was "growth hacker," then "marketing technologist," now "GTM engineer" and "marketing engineer." Honestly, I don't think the title will survive. The skill will.

How is a marketing engineer different from a GTM engineer or marketing ops?

Marketing ops runs the machine. A GTM engineer builds the machine for sales. A marketing engineer builds the machine for attention. The three roles overlap, and the job titles are still a mess, so here's how I draw the lines.

Marketing ops keeps the existing stack working: clean CRM data, campaign setup, attribution reports. Success looks like data you can trust.

GTM engineers build the revenue engine for sales: enrichment, outbound sequences, lead routing. Success looks like meetings booked and pipeline.

Marketing engineers build the systems that create and distribute marketing: listening tools, content pipelines, community monitors, AI search presence, agents. Success looks like reach, citations in AI answers, and how much one person can run.

Think of a restaurant. Marketing ops keeps the kitchen clean and the orders flowing. The GTM engineer builds the delivery system that gets food to customers who already ordered. The marketing engineer builds the thing that makes people walk past and want to come in. Usually a marketing engineer combines the roles of a whole team of specialists. I like to call them the king of agents and systems.

What does a marketing engineer actually build?

Everything I've built fits into one of five layers: Listen, Understand, Make, Ship and Learn. I call it the marketing engineering loop, because what you learn at the end changes what you listen for at the start.

1. Listen. Where is the conversation happening right now? Keyword monitors, social radars, forum and review scrapers.
2. Understand. What does it mean for us? LLM-summarized research, pain-point maps, competitor tracking.
3. Make. What do we say, and in what form? Content, video and voice pipelines.
4. Ship. Where does it need to live to be found? Multi-agent systems that ship content across channels, owned communities, presence in AI search.
5. Learn. What worked, and why? Performance logs, experiment tracking, and feedback back into Listen.

Here's the thing most teams get wrong. They start with Make, because "AI writes posts" is the obvious use case. But content volume was never the bottleneck. In my experience the biggest gains sit in Listen and Learn. Listening finds the moments worth acting on. Learning tells you which of your actions actually mattered.

What should marketers automate first?

Automate tasks that repeat often and are safe to get wrong. For tasks that repeat often but happen in public, let the machine draft and keep a human on the send button. I call this the frequency × risk matrix.

The most common mistake I see is automating whatever is easiest to automate. That's how you end up with a bot that confidently replies to a grieving customer with a discount code. Two questions work much better: how often does this task repeat, and how bad is it if the machine gets it wrong in front of people? The answer is a split: the machine finds and drafts, a person approves and sends.

Five marketing engineering systems I built, one per layer

Theory is nice, but you came here for the real stuff. Here's one system for each layer of the loop, with what it taught me.

Listen: a client-finding radar on Threads

Situation. I wanted to find advisory clients without cold pitching. I hate cold pitching, and so does everyone receiving it. The best moment to talk to a founder is when they post "we just raised" or complain that growth has stalled. Those posts disappear from the feed in hours.

What I built. A radar that searches Threads for pain-point phrases in English and trigger phrases in Ukrainian, like "we raised" or "we got our first revenue." For every match it drafts a reply and pings me in Telegram or email. I read it, rewrite it if needed, and post it myself. The system never posts on its own, and nobody gets a pitch in the first message.

What it taught me. Listening is the cheapest layer to automate and the most valuable. Timing beats copy.

Understand: market research that turns into product

Situation. At an AI workflow automation platform, we were selling agents to small service businesses in seven verticals. To sell to them, we needed to know which tools they already use and what actually makes them swear at their phones.

What I built. A research system that produces three things per vertical: an integration map of the software they already use, a pain-point file built from scraped Reddit threads and classified by an LLM, and ready-made agent templates that solve the top pain with the tools they already have. The research didn't just shape the message. It turned into the product itself: the templates.

What changed. Cost per lead dropped by 80%. Instead of guessing, we shipped 7 agent templates, one per vertical, each built around a pain owners described in their own words. Our one-size-fits-all pitch was gone.

What it