Margin Notes
A weekly publication on AI for marketing — annotated, opinionated, written from the practice.
Most of what is written about AI for marketing comes from two places. Vendors, whose conclusion was decided before the first sentence. And commentators, who have never had to introduce a tool to a sceptical team on a Monday morning and still have the team's trust on Friday.
This publication is the third place. I lead marketing for a deep-tech company in Paris and consult for B2B teams introducing AI to their work. Every week something works, something fails, and something I believed three months ago stops being true. Writing it down once a week is the discipline that turns those weeks into a body of knowledge — mine, and if it's useful, yours.
The name
Marginalia is what expert readers leave in the margins of other people's books. The annotations are where the reader argues with the author, links to the better paragraph two chapters back, marks the sentence worth coming back to. Margin Notes is that, applied to AI marketing — and to my own past pieces. Cite, disagree, revise, return.
"Margin" has the second meaning too. Marketing departments are measured on it; the publication keeps it in view. The interesting problems live where AI actually moves margin — without spending brand voice, customer trust, or the team's confidence in its own craft to get there. The body of work below is about exactly that line.
The cadence
One piece, every Tuesday. Not more. I've made the case against daily content elsewhere, and this publication applies it to itself. Fifty-two pieces a year is a quarter of what the algorithm wants and roughly four times what most people can sustain with quality. It's the deliberate middle.
The five formats
Every piece is one of five shapes, so you know what you're getting before you click:
-
Field notes
What actually happened this week — what worked, what didn't, with enough detail that you could repeat or avoid it.
-
Opinion
A take I'll defend. Sometimes a disagreement with something popular. Always falsifiable.
-
Breakdown
A brand, team, or tool taken apart on the page. What they did, why it worked, what you can lift.
-
Playbook
A procedure your team can run on Monday. Steps, time estimates, and where it goes wrong.
-
Reading week
What others wrote, annotated. I'll tell you why each piece matters and where I think it's wrong. Agreement not guaranteed.
There is a sixth, quieter shape: notes — evergreen definitions of the vocabulary that runs through everything else. Those aren't weekly; they appear when a term needs a permanent home.
The rules I write by
Cite generously. If someone else said it better, you'll get their link, not my paraphrase. Disagreement gets the same courtesy: named, linked, and argued with on the merits.
Revise publicly. When I change my mind about something I published, the new piece says so and links the old one. The old piece stays up. A publication you can trust is one with a visible memory.
Pass the eighteen-month test. Before anything ships, one question: will this still be true in eighteen months? Tool reviews date; principles don't. The mix here leans hard toward the second.
Who writes this
Eduardo de la Espriella — Eddie. Marketing Team Lead at Outsight in Paris, where I run global marketing for a 3D-LiDAR spatial intelligence platform. MSc in AI for Marketing Strategy. Eleven-plus years across Europe, Latin America, and the US. The full record is on the CV; the practical version of this publication's ideas is the free AI marketing audit.
Subscribe by RSS, or get the monthly digest by email. If you only want to argue with one piece, write to me — disagreement is the point.
Common questions
Why "Margin Notes"?
Two meanings, both work. Marginalia is what expert readers leave in the margins of other people's books — annotations, links, arguments. The publication does that to AI marketing, including to its own past pieces. The second reading is operational: marketing departments are measured on margin, and the interesting problems live where AI actually moves it. Both readings are the point.
How often does it publish?
Once a week, every Tuesday at 09:00 CET. Fifty-two pieces a year — about a quarter of what the algorithm wants, and roughly four times what most creators sustain at quality. Slower than the feed, faster than a monthly column.
Who is it for?
Marketing leaders, CMOs, founders, and senior marketers at B2B technical companies — deep tech, SaaS, industrial AI — who need to introduce AI to existing teams without losing brand voice, customer trust, or the team's confidence in its own craft.
How do I subscribe?
Three ways. RSS — full-text articles, no email required. The monthly digest by email, pointing to the four most recent pieces. Or follow on LinkedIn — every piece's deck posts there on Tuesday morning.
What are the editorial rules?
Cite generously — link the original, never paraphrase-and-bury. Revise publicly — when I change my mind about something I published, the new piece says so and links the old one; the old piece stays up. Pass the eighteen-month test — if a piece won't still be true in 18 months, don't publish it.