Most AI-generated GTM campaigns sound generic for a simple reason.
The model is doing exactly what you asked.
You gave it generic context.
"We sell to B2B SaaS."
"Our ICP is VP Sales."
"We help teams grow pipeline."
"Mention AI transformation."
Of course the output sounds like every other campaign.
Good GTM campaigns need sharper inputs:
1) the buyer's actual job
2) the trigger that makes this urgent
3) the current workaround
4) the competitor or alternative in their head
5) the specific use case
6) the proof you can point to
7) the reason this message should land now
That is not prompt engineering.
That is context engineering.
The prompt matters.
But the context underneath matters more.
If your GTM context is scattered, your AI campaigns will average out into slop.
If your GTM context is structured, the same tools suddenly get much more useful.
I finally became the type of guy that goes into flow-mode-adhd-pro-max and launches 5 parallel agents to run content, GTM, video editing and engineering work 😭🥹
Introducing the 4th edition of the State of AI ✨ Latin America.
In November 2025, AI experienced its Agentic Big Bang. New tools showed that AI could do more than answer questions. They could take action.
Today, we’re beginning to see how AI agents are transforming organizations across Latin America.
Built with insights from 420+ founders, operators, investors, and executives, this is our most comprehensive report yet. Report in thread.
In partnership with @ElevenLabs@facesdotapp@federicoantoni@pimepardo@lavca_org
The new vibe maximizing lifestyle is crazy:
Every 4 hrs
Plan session
Connecting to claude pro
Working for 30 minutes
Running out of tokens
Get a long break, recharge
Repeating the cycle
Bug Bounty Switzerland raised CHF 12m - I was bold enough to break down their GTM intelligence live in a Clay event in Zurich, in front of their GTME.
(I thought he'd kill me) - but he was actually surprised about it.
Talking about Lucas Maliczak🐐, the GTM Engineer behind it, who organized an amazing Clay GTM event in Zurich.
He gave a keynote on Claude Code and Clay - then after the talk, he invited me on stage to show how I do it with the Appsaavy GTM Engine.
Here's what happened:
I opened Claude Code, walked through the framework live: it breaks a company into GTM entities, runs deep research on each, then builds a structured, editable database inside Zite.
What makes it powerful is now turning that raw database into a live, team-facing application that serves as a source of truth for GTM intelligence.
It's fully connectable back with Claude Code, automations or any agent.
Then I showed how the participants could pretty much skip the hustle of figuring it out from scratch and let Hermes, my GTM agent, runs the whole thing for them with https://t.co/XufunUlo2F
We used Lucas's company as the demo. I typed in the domain, hit go.
5 minutes later: the full database was created with real data from the company.
1 click to "claim my app" led to a fully-built GTM dashboard:
→ Buying triggers
→ value propositions
→ ICP breakdown
→ competitors
→ segment playbooks
The kind of output that takes a GTM consultant multiple interviews and days to build.
All of it live on stage.
All of it editable and ready to kickstart a GTM project with AI agents, now with the right context.
Want to see how accurate it is for yours?
If you want to run it on your own company, it's free at https://t.co/XufunUlo2F
Most GTM engineers are stuck at level 1. There are 3.
Level 1: Workflow Builder You automate one task at a time. Each automation is its own island. Nothing feeds into anything else.
Level 2: Systems Integrator You stop building tasks and start building funnels. A signal fires. The next step runs automatically. Workflows talk to each other. The pipeline moves without you in it.
Level 3: Revenue Architect You design the full engine. You think in data models, attribution loops, and feedback cycles. You don't ask "how do I automate this?" You ask "what does the system need to know to make the right decision?"
At this level you don't run campaigns. You build infrastructure that runs them.
The jump from L1 to L3 isn't a tool problem. It's the shift from task thinking to system thinking.
GTM teams are drowning in tools.
500+ platforms. Overlapping features, overlapping data points.
Zero clarity on what actually works together.
CMOs and CROs are building stacks blind, hoping they picked the right pieces.
We're building the map, the universe.
A live observatory of GTM tools, integrations, and real stack architectures from teams that are actually shipping.
No vendor listicles. Just the truth about what connects, what breaks, and what works.
Coming soon 🔭
Después de un año fuera de Colombia, estoy de vuelta.
Salí con un propósito y un anillo.
Desde entonces:
- Me casé
- Viajé por 5 países como nómada
- Me mudé a Barcelona
- Conseguí un trabajo alineado con mi propósito y habilidades
- Comencé una nueva empresa
- Estoy creando productos de forma independiente
- Vivo entre España y Suiza
- Pasé un mes en Suiza
- Hice amigos de todo el mundo
- Estoy planeando mi boda
Y vuelvo a Colombia para casarme con la mujer de mis sueños.
Cuando se trata de lead generation, todos están obsesionados por el tamaño de la lista - ese es el problema.
He visto a equipos a segmentar listas desde 40.000 hasta medio millón de empresas. En muchos casos las listas las hacen con filtros básicos, sin mucha lógica.
Al final, terminan quemando créditos de Clay. Y peor aún, con resultados mediocres.
Yo aprendí a verlo diferente.
La verdadera pregunta no es cuántas empresas tienes, es cuál es el propósito de cada una en tu lista.
La segmentación es específica al problema que solucionas. Un CRO muy duro que fue mi jefe lo dijo mejor: "the list is the message".
Y tiene toda la razón.
Cuando construyo una lista de mercado (TAM) en Clay, pienso primero en esto: cada tabla necesita arquitectura.
No puedes meter datos y esperar que funcione. Tienes que planear de donde puedes sacar la data y como la vas a segmentar la data, desde el inicio.
Ejemplo real: tu cliente vende AI Agents para Customer Support.
No importa si tienes 50.000 empresas listadas. Lo que importa es identificar cuáles tienen mínimo tres personas en ese departamento.
Eso es específico. Eso es relevante para tu GTM.
A diferencia de hablarle a todos, le hablas a los que tienen *sintomas* del problema que tu solucionas.
Y esos sintomas no siempre son faciles de encontrar, pero los agentes de IA suelen ayudar. Lo importante es que no le pido a los agentes de IA que decidan por mí.
Les pido que saquen evidencia. Datos concretos. Y luego yo decido qué hacer.
Algunos usan una fuente de datos. Otros combinan múltiples. Todo depende de tu contexto de campaña.
No es rápido. Pero hace la diferencia.
La segmentación es el 90% del trabajo. Si tienes el público correcto, incluso un mensaje promedio convierte.
Si tienes el público incorrecto, nada te salva.
Optimiza por relevancia, no por creditos o volumen.
Y en tu empresa, ¿la lista es grande o es buena?
The best lists come from collaboration between someone who knows the data tools and someone who knows the market.
Not from back-to-back defensive conversations.
About the pic: this was me on the forest that day - much needed break.😮💨
I spent 40 minutes explaining to a marketer why his Clay list wasn't "broken."
He found 3 people on Sales Nav that weren't in the list I built.
So obviously, my work was wrong. 🧵
Your GTM Engineer verifies data quality. They build logic to filter and enrich.
They can't read your mind about edge cases you or the SDRs have seen in your many years selling to this market.
If you want a highly targeted list that matches your exact expectations, get involved early.
Bring your knowledge to the table.
Don't wait until it's done to play "gotcha" when deadline is close.