Forwarded this email? Subscribe here for more.
👋 Hey there, I'm Nick. Each week, I share field notes from the people building the future of marketing. For more: LinkedIn | The Marketing Engineer | Profound University.
Matt Swulinski is one of the most brilliant marketing engineers I've spoken with.
The systems he has built at Wispr Flow are world class. He runs real growth workflows through Claude Code, from insertion orders to newsletter copy to internal context management. The setup is technical, opinionated, and clearly built by someone who has spent a lot of time turning messy marketing work into repeatable systems.
The lesson I took from him was more specific than "use more AI":
The hardest jump in marketing engineering is moving from single-player tools to multiplayer systems.
It is easy to build for yourself. You know the edge cases. You remember why the prompt is shaped that way. You know where the context lives. You can tolerate a little weirdness because you built the thing.
Building for a team is different. Now you have real users. They do not have the same context. They do not know why you made each decision. They need the system to explain itself, survive bad inputs, handle permissions, stay current, and keep working when you are not standing next to it.
That is the bottleneck every marketing engineer will hit.
1. Personal leverage is the first win. Team leverage is the real test.
Matt's clearest advice was simple:
The number one piece of advice I would give is design the tools and the systems you build today to be multiplayer.
This is where the work changes. A personal agent workflow can be powerful while still being fragile. A team workflow needs a shared source of truth, clear interfaces, feedback loops, auth, permissions, and enough structure that someone else can use it without inheriting the builder's brain.
That is the marketing engineering bar.
The phrase "AI-native" already feels too broad. Most marketers can open ChatGPT, Claude, or a dozen SaaS tools with AI features. The scarce skill is taking that raw capability and adding an engineering layer: systems, state, shared context, monitoring, and repeatability.
Matt has clearly crossed that line.
2. The terminal is becoming the marketing control plane.
The thing that surprised me most was the depth of Matt's technical workflow.
His primary interface is the terminal. The work runs closer to files, code, and skills than to another web UI. He is orchestrating work through Claude Code, skills, files, and feedback loops. That distinction matters.
I think this is a preview of where marketing work is going.
The next generation of marketing operators will control fleets of agents. Some will write copy. Some will inspect data. Some will update context. Some will route tasks to other systems. The marketer's job shifts toward designing the workflow, maintaining the source of truth, and giving the system feedback when it misses.
Matt is building that standard right now.
This is also why the terminal matters. It is a control surface for orchestration. A SaaS dashboard gives you a workflow someone else imagined. A terminal-based system lets you compose your own.
That is uncomfortable for a lot of marketers. It also creates a huge opening for the ones willing to learn.
My advice: learn how to use the terminal. Ask Claude to teach you.
3. The workflow to steal is the one that turns messy input into usable output.
The most tactical thing I want to steal from Matt is his newsletter sponsorship workflow.
The old version of that process is painful: contracts, insertion orders, rate cards, audience notes, deliverables, copy requirements, links, reporting, and a bunch of coordination over email. Every partner has a slightly different format. Every placement has slightly different constraints.
Matt built a system that can take an insertion order and turn it into the downstream marketing work. It understands the partner, the constraints, the copy needs, and the internal context.
That is exactly the kind of workflow marketing teams should be looking for.
The best AI workflows start with a real business object: an insertion order, a sales call, a customer research doc, a launch brief, a content inventory, a support ticket. Then the system turns that messy object into structured work.
That is the difference between playing with AI and building leverage.
What I am taking from this episode
Three things.
First, I need to judge marketing systems by whether they can survive handoff. If only the builder can use it, it is still a personal tool.
Second, I want more of my own workflows to use files and shared context as the durable layer. Matt's point about avoiding hard-coded facts in skills is exactly right. The system should read from a source of truth instead of carrying stale facts around forever.
Third, the best marketing engineers are going to look more technical than most teams expect. Many will come from marketing, but they will be comfortable operating closer to code, files, terminals, agents, and APIs.
That is where the work is moving.
And Matt is one of the clearest examples I have seen.
Watch or listen
If that's you, or you want it to be, stick around. Hit reply and tell me what you're building right now. I read everything.
See you next time,
Nick