Ygor Fonseca

Founder & Systems Lead at Leanboat | AI-native growth for DTC and B2B SaaS | I build the systems underneath the work

[Leanboat @leanboat_agency]

If your AI marketing still feels like a chat window, the gap is not your model. The function it was supposed to run was never written down.

Read: https://t.co/0orR1EA79t

[Ygor Fonseca @fonseca_ygor]

Two brands, same budget, five years. One's marketing compounds; the other starts from zero every quarter. The gap isn't spend or talent — it's whether you define your entities once and propagate them intact, or reinvent them each campaign.

Read: https://t.co/c8bmp7j9i5

[Ygor Fonseca @fonseca_ygor]

Type your company name into ChatGPT. You get three sentences about what you do and which competitor you're a worse version of. Nobody wrote them — a model did, from your site and a 2023 directory.

Your About page isn't copy. It's training data.

Read: https://t.co/puK68buYwG

[Ygor Fonseca @fonseca_ygor]

When an AI engagement ends, the one thing you own outright is the memory layer — the hard-won, per-brand lessons a wiki never holds: what got walked back, what lost, and why. The methods travel; that never does.

Read: https://t.co/xmeMHplCJJ

[Ygor Fonseca @fonseca_ygor]

A number you look at but never act on is a scoreboard, not a feedback loop. Most "AI content" engines have a dashboard and call that learning. The fix is mechanical: the output of measurement has to be an edit to the next plan.

Read: https://t.co/KLffkRunPs

[Ygor Fonseca @fonseca_ygor]

One AI check isn't a gate. A second look only protects you if it can fail differently from the first — a checker that shares the writer's blind spots adds confidence, not safety. And confidence is the one that ships.
Read: https://t.co/Cpmc5tSMjl

[Ygor Fonseca @fonseca_ygor]

The demo promises campaigns by Friday. Then week one of the AI install isn't about agents at all — it's one customer scattered across five tools that don't agree she's the same person. Stitch her first, or the agent just runs fast on wrong facts.

Read: https://t.co/eGKCUD41rw

[Ygor Fonseca @fonseca_ygor]

Everyone treats AI marketing as a model problem. It's a codification problem. Service-as-Software reached legal and insurance first because their work was already legible. Marketing's messy middle lives in DMs and Looms — no model reads that.
Read: https://t.co/9KBo2qYAUh

[Ygor Fonseca @fonseca_ygor]

AI search cites original evidence — data, quotes, first-hand expertise it can't find elsewhere.

We sampled 9 GEO/AEO vendors: 7 sell mention-tracking, none sells the citable thing.

The dashboard measures citations. It can't manufacture them.

Read: https://t.co/Aa0zapz0Bu

[Ygor Fonseca @fonseca_ygor]

Most agencies keep their method in a senior person's head or a slide deck. One walks out the door; the other can't run. The third option: a library of small instruction files that do the work — and compound every engagement.

Read: https://t.co/zU9FCIhi0i

[Ygor Fonseca @fonseca_ygor]

Most agencies sell content as a calendar of posts — which is why the work never compounds. Content is really a 5-layer operating model. Staff only execution, and headcount is your ceiling.

Read: https://t.co/hVNgVx9vF0

[Ygor Fonseca @fonseca_ygor]

Two AI-services firms pitch the same six-month build for the same price. One wins. The tell isn't the deck — it's whether they can name the six primitives, say workflow or agent, and show you where the kill switch sits.

Read: https://t.co/VibvJeCNkA

[Ygor Fonseca @fonseca_ygor]

An AI-native agency ships a brand-install, not a team.

Same senior lead, same AI in the loop — but one hands you people, the other hands you a system that keeps working after they leave. Everything else is org chart.

Read: https://t.co/1Jdai4nEct

[Ygor Fonseca @fonseca_ygor]

Same self-improving-loop framework, same AI tools, two teams. Three months later one compounds and one doesn't. The gap is the install: a folder of markdown files the AI reads and the team versions like code — not a SaaS product.

Read: https://t.co/3Ie3jG9Ii0

[Ygor Fonseca @fonseca_ygor]

The real AI risk in marketing isn't hallucination. It's unreviewed AI output quietly becoming organizational truth.

The discipline teams are converging on: a pull-request culture for knowledge — nothing merges without a reviewer.

Read: https://t.co/OKpwcAub34

[Ygor Fonseca @fonseca_ygor]

Every AI-native services firm names its operating method — Plan/Work/Assess/Compound, AOPs. The naming is the moat. Can't name yours in two words? That's not a marketing gap. It's an operations gap. The marketing is downstream.

Read: https://t.co/vGiO3RQ4zt

[Ygor Fonseca @fonseca_ygor]

If your AI marketing feels like a chat window, the function was never written down. The model isn't the gap—codification is. Name the stages, own the gates, close the loop. The model is downstream of all of it.

Read: https://t.co/kb9bKEcsVI

[Ygor Fonseca @fonseca_ygor]

Most “AI agency” work is a campaign: they run the AI, you see outputs, and when the engagement ends, the system ends. The right shape leaves a nervous system behind — in your repo, on your data, with your keys. An install, not a deliverable.

Read: https://t.co/yCuV3UpKue

[Ygor Fonseca @fonseca_ygor]

Org charts were never the goal. They are a 2,000-year workaround for one hard limit: a human can coordinate 3-8 people directly, no more. Every experiment to flatten them reverted. What changes when coordination stops being a human problem?

Read: https://t.co/rI3q5QkOJZ

[Ygor Fonseca @fonseca_ygor]

In one quarter, AI-native services went from thesis to roster: 6 verticals, $20B+ in value. Same shape each time — AI reads, a human owns the judgment, sold productized not by the hour. Marketing is the last big vertical still selling hours.

Read: https://t.co/MX03WaIGrc