An Engineer’s View on Running a Newsletter Like a System, Not a Marketing Campaign

Julien Coche
35 minutes ago
7 min read

Content marketing has a reputation problem.

For many founders, operators and technically minded leaders, the phrase sounds soft. It suggests blog posts written for search engines, social posts written for algorithms, and newsletters sent because someone in marketing said, "We need to stay top of mind."

That skepticism is often justified. Much content marketing is not very useful. For a time it was basically articles stuffed with keywords to win Search Engine Optimization (SEO), produced because a calendar had empty slots or a team had output targets. The result is predictable: more content, more noise and less trust.

However, as you may have guessed, content marketing has (surprisingly) taken a bit of my time lately. At Agentic Foundry, we spend a lot of time on newsletters, so I decided to delve into this branch of marketing to understand what it was all about. Here are my notes and observations.

TL;DR. A newsletter becomes far more credible when you stop treating it as marketing output and start treating it as a reliable communication system: owned, instrumented, maintained and trusted over time. A good business newsletter is not an email. It is a service line between a business and its market that turns internal expertise into external trust, repeatedly, and building one is at least as much an engineering problem as a marketing one.

This piece is about the system itself, the spec, cadence, pipeline and deliverability a real newsletter strategy runs on. The payoff, why a newsletter becomes your most durable owned audience, is the subject of the companion piece, Why a Newsletter Is Your Most Durable Owned Audience.

This is infrastructure, not a campaign

A campaign is a discrete set of events, with a start, an end, a goal and a deliverable. It asks: what do we announce this week? A system manages a continuous flow, with inputs, dependencies, failure modes and maintenance. It asks a different question: how do we stay useful, visible and trusted every week?

Most companies say they want the second outcome but operate with the first mindset. They send when there is something to push, and from the reader's point of view the newsletter becomes a series of interruptions organized around the company's needs. A system-oriented newsletter starts elsewhere, with the reader's ongoing relationship to the business: what is worth receiving from this company regularly?

This is really Seth Godin's conceot of permission marketing, setting the bar at communication that is anticipated, personal and relevant, and permission is not a permanent asset. It is renewed or degraded every time you appear in the inbox. So the useful question is not "how often can we send?" I'd argue it is "are we increasing or decreasing trust every time we appear?"

That key idea shifts the goal from volume to usefulness. A newsletter is not a megaphone. It is a recurring product that must earn its place in the reader's routine.

Your newsletter needs a specification

Many bad newsletters are not bad because the writing is terrible. They are bad because the system has no clear specification. Who is it for? What job does it do for that audience? What is the editorial promise? What cadence can the company sustain? What belongs in it, and what does not? What counts as success?

Without answers, the newsletter becomes a dumping ground where company news and promotional blurbs get poured into the same container. This is where the engineering lens appears. A newsletter is, in a real sense, a product a company offers its audience. It may be free, but it still consumes attention, creates expectations and needs a reason to exist. No serious team would ship software with no requirements, no release schedule and no telemetry, yet many companies run their newsletters exactly that way.

So write a spec. Not a 40-page document, but a clear definition: who it serves, the promise, the recurring value, the editorial boundaries, the tone, the cadence, the action the reader is invited to take, and what you will measure and deliberately ignore. Without one, the newsletter gets optimized around internal convenience, because the easiest thing to send is whatever the company already wants to say, and the hardest is what the reader actually finds useful.

Cadence is a service-level commitment

The research and the practitioner advice both support a regular cadence, which makes sense, because trust compounds through repeated, recognizable interactions. But the nuance matters: regular mediocrity simply trains people to ignore you on schedule.

Consistency is valuable only when the thing delivered has value. The engineering analogy is uptime, and uptime for a bad product creates no loyalty. So treat the cadence like a service-level commitment: frequent enough to build memory, realistic enough to maintain quality and stable enough that the reader knows what they are entering. For some businesses weekly is right; for others, twice a month is more honest. The answer is rarely "send more." It is to send at a rhythm where you can keep both the cadence promise and the editorial promise.

This is another place campaign thinking causes problems. Campaigns tolerate spikes; systems need sustainable throughput, which is where automation earns its place. If a newsletter depends on last-minute heroics, it eventually fails: quality drops, the voice drifts, approvals bottleneck and every issue becomes an emergency.

A good system reduces drama and creates a repeatable path from idea to send. I want to insist that the point is not to remove creativity. It is to make creativity reliable.

Focus on the production pipeline

Once a newsletter is treated as a system, the production pipeline comes into focus.

There is an input layer (customer questions, sales calls, support tickets, product updates, internal expertise), a selection layer (which ideas are worth sending, to whom, and when), a shaping layer (angle, structure, evidence, tone, call to action), a quality layer (editing, fact-checking, link-checking, rendering tests, compliance, deliverability checks), a distribution layer (segmentation, scheduling, authentication, list hygiene) and an observability layer (clicks, replies, conversions, unsubscribes, downstream sales influence). And there is a learning layer: what should change next time.

This sounds more complicated than "write an email," but it is what good teams already do. When the pipeline is undefined, every send is improvised and the key learning loop is missing.

For a small business the pipeline can be lightweight: a monthly review of customer conversations, a short editorial calendar, a reusable structure and a clear approval owner. For larger or PE-backed businesses it needs governance, because multiple teams contribute, several locations hold local knowledge, legal or brand review may apply and customer data lives across systems.

There the bar rises from "can we write a good newsletter?" to "can we repeatedly produce useful, compliant, on-brand communication without turning it into generic corporate paste?" Interestingly, it remains a system question.

AI changes production, not the trust problem

AI changes newsletter production in obvious ways. It helps with most of the steps, from ideation and drafting to summarizing customer interviews and adapting one idea across segments, and it lets a small team produce like a larger one. But sadly, AI does not solve the problem that matters most: trust. In fact, it can make things worse.

When many companies use similar tools, similar prompts and similar source material, the output becomes more polished and less distinct. Automation increases sameness faster than it increases usefulness, the same dynamic we wrote about in The Average Machine: AI raises the floor on quality and flattens what makes a voice distinct.

This is why I doubt the winning model is "AI writes the newsletter." It is also not a human rubber-stamping, which just becomes a bottleneck wearing a quality-control badge. The better model is human-on-the-loop: you encode your judgment into the system's rules and guardrails instead of re-applying it issue by issue. Humans define the promise, insight lense and source material. AI drafts and transforms inside the guardrails.

Both actors have roles. Humans edit for accuracy, specificity and voice, approve anything sensitive and step in by exception. AI makes the system faster. It cannot, by itself, make the newsletter worth receiving.

This system-not-campaign view has a name. It is the core of what we call Agentic Content Marketing: content produced and improved as a coordinated system rather than shipped as one-off campaigns. Full disclosure, it is also the thesis behind Marabel, the Agentic Content Engine we are building at Agentic Foundry to run a newsletter exactly this way, with humans on the loop. This is the difference between a newsletter that drains the team and one that compounds.

Deliverability is the infrastructure layer

A newsletter does not matter if it does not reach the inbox, and deliverability has gone from technical footnote to core infrastructure. Google's sender guidelines require authentication through SPF, DKIM and DMARC, ask senders to keep spam complaints below 0.3% (with 0.1% as the real target) and require one-click unsubscribe for anyone sending more than 5,000 messages a day to Gmail. As of late 2025, Google moved from soft deferrals to outright rejections for mail that does not comply.

This is healthy. It treats distribution as a trust system rather than a neutral pipe, rewarding senders who behave predictably and penalizing those who look risky or unwanted. It also makes content discipline part of the infrastructure. If people did not sign up, do not send. If a segment is disengaged, sending more will not repair it.

A clean, one-click unsubscribe is not a marketing failure. It is a sign of a healthy system. Deliverability is to newsletters what uptime is to software: invisible when it works, existential when it fails.

What this means for building one

None of this turns a newsletter into a NASA program. It means designing the system around the work and then letting it run. Write the spec, so it knows who it serves and what it promises. Pick a cadence you can hold. Build a real pipeline from input to send, so no issue becomes an emergency. Put AI inside the guardrails and keep humans on the angle, the judgment and the final call. Treat deliverability as infrastructure.

Do that, and the newsletter stops being a weekly act of corporate self-expression and becomes a system that reliably turns expertise into trust. The harder question is what that trust is worth, and why a newsletter is the most durable audience a business can own. We’ll cover that in an upcoming companion piece, Why a Newsletter Is Your Most Durable Owned Audience.

Agentic Foundry: AI For Real-World Results

Learn how agentic AI boosts productivity, speeds decisions and drives growth

— while always keeping you in the loop.