Hay gente acá vendiendo herramientas de outbound como la solución al GTM.

→ Instantly no te resuelve el GTM.
→ Heyreach no te resuelve el GTM.
→ Ese AI SDR tampoco.

El mercado está saturado de campañas en frío mal ejecutadas.

Y la educación que ofrecen suele prometer: 
"En 20 minutos enviarás una campaña"
"Solo necesitas una herramienta"

Algunos se emocionan y ni siquiera necesitan outbound.

Se les olvida que es solo uno de muchos canales de GTM, no la solución universal.

La mayoría cree que enviar buenos cold emails es cuestión de:

→ Armar una lista
→ Escribir un copy
→ Enviar una secuencia

Pero la realidad es otra.

Outbound efectivo requiere pensar en sistemas, en señales de compra, en llegarle a la persona correcta, en el momento correcto, con el contexto correcto.

Ahí es donde entra el "GTM Engineering":

Una disciplina que va mucho más allá de un software.

Es arquitectura, campañas, datos, señales y diseño de automatizaciones.

El mundo no necesita más gente lanzando campañas al azar.

El mercado ya está saturado.

Eso es receta segura para conversiones mediocres.

Lo que realmente mueve la aguja:

→ Las señales que activan el outreach
→ El canal y formato adecuados
→ Un mensaje relevante atado a la propuesta de valor
→ Una arquitectura que respalde ese workflow

La magia pasa antes de abrir cualquier herramienta.

Saber usar Instantly o HeyReach es valioso. Saber Clay es aun mas valioso.

Pero vender un software como "la solución al outbound" es precisamente la razón por la que la mayoría de campañas fallan.

Outbound sigue siendo poderoso.

Pero solo cuando está construido desde estrategia de ventas, no desde la urgencia.

Outbound Marketing Misconceptions

Rant monday: @clay_gtm is worth a billion dollars and has engineered their users into wasting money every time records go stale.

Either way, you're paying twice.

I'll give you a reason to reconsider before investing heavily in Clay: 

Records going stale and forcing you to re run them is my opinion the #1 biggest flaw and biggest pain with Clay.

I thought I had learned how to avoid stale records (with a few rules like not changing input columns), then this happened today.

Clay's infrastructure is a magic box so fragile that any little change you make can mysteriously break your data tables and make your records unusable for further enrichment... forcing you to double-spend every time that happens.

It's so annoying that other builders have shared they've moved software because of it.

If you know me, you know I'm a fan of this product (hell I got a job because I know it), but for a BILLION dollar company selling enterprise tiers at the price of a full time person, you'd expect something more stable and engineered for success.

1.  I don't see enough Clay experts calling them out on their broken pieces.

2. Am I the only one struggling with this?

3. Why is this a thing! In which other automation software or spreadsheet does changing a configuration lead to your records being unusable?

That's my rant, good monday.

A Rant on Clay's Stale Records

Gemini 3.0 is already making a business case in our GTM.

It's a big change for anyone processing thousands of records: running research at scale, segmentation, or vertical-specific TAM mapping.

I’ve been building a data pipeline to identify all potential TAM companies in the energy sector.

Our segmentation criteria requires crawling company websites to extract structured details about their renewable projects.

The challenge?

👉 Crawling and returninig structured research at scale is expensive.

For someone taking a shortcut, enriching 30k companies with claygent (3credits per row) is already 90k credits. That means getting a $1.5k subscription from Clay.

So, running a tables with across ~50k companies can easily become an expensive problem.

Until now, I've used cost-efficient “waterfalls” that looked like this:

↳ First-layer research → Perplexity $
 ↳If it passes the first check → Claygent $$
  ↳If still needs a second source → Exa Research $$$

Perplexity can often returns false positives or mixes companies with similar names.

So today, I tested this research with Gemini 3.0 Pro from Google AI Studio.

And the results were insane:

→ More accurate research
→ Cleaner structured outputs
→ Almost no hallucinations

I found research depth is still limited to surface-level website, news and media links, so it didn't go as deep as Exa or Claygent, but still brought in reliable results, very useful as a first layer.

And the cost…

👉 $0.0031 for a 50 second crawling sessions with deep reasoning and structured outputs.

👉 That's aproximately 30 cents per 100 records.

Read that again.

We’re now at a point where you can:

→ Crawl thousands of websites

 → Extract structured signals

 → Score, segment, and enrich

 → Power entire TAM models

 → …for a few dolars instead of thousands.

I don’t think people understand how big this will be.

I’m not even sure I fully do yet.

But one thing is clear:

Vertical GTM automation is entering a completely new cost frontier.

And Gemini just unlocked it. 

Gemini 3.0: Scaling TAM Research

I think we’re entering the era of micro GTM apps.

Tiny, vertically-specific tools built from winning:
→ Clay workflows
→ Relay flows
→ Workflows
→ Lindy automations

Not “AI startups.”
Not “platform plays.”

Just tiny functional apps that:
→ have already been proved in sales environments
→ solve one painful GTM job
→ are built in hours
→ and can stack like Lego blocks

Example apps:
- Tam mapping for boring industry
- Monitor SAAS tool users
- Create ads countering competitor curreny ads
- Niche market research agents

This will happen.
And many more like it.

The tools are here to turn your automations into micro apps.

Distribution is already your expertise.

If you’re a GTM engineer today, you’re sitting on a goldmine of productizable workflows.

The question is: will you turn them into apps before someone else does?

Micro GTM App Revolution

The headlines: Google is coming for N8N, Relay, ChatGPT, Zapier and your job! 

Meanwhile the product:
__

Gave a quick shot at Google's new native flows.

And yet again, the AI-hype moved faster than the actual workflow builder.

In reality:

Yes, Google launched a new workflow builder.

No, it's not able to execute on a simple pre-call automation use-case.

It took me 5 minutes and two attempts to confirm this product is nowhere near replacing any serious AI workflow or automation tool.

BUT…

For everyday builders?
For small wins inside Google Workspace?
For people who don’t want to commit to a new tool or subscription?

👉 This can be a solid starting point.

The UI is intuitive.
The blocks feel familiar, with the tools people already use.
And it lowers the barrier for “automation beginners” who already live in Gmail, Sheets, and Calendar.

If Google keeps pushing, this could get a lot of people past the
“new tool fatigue” + “another subscription?” objections…
and finally into experimenting with productivity automation.

But for now?

If you’re doing anything beyond trivial workflows,
Relay, Zapier, Make, n8n, Lindy and friends remain miles ahead.

✦ Have you gotten any wins from this tool?

Google's New Workflow Builder

Found this video from back in the early days of ChatGPT when I was experimenting with what @Bardeenai + OpenAI API could do.

The use-case: making essays from thought leaders using their recently scraped tweets (recorded in early 2023).

Nowadays, this seems nonsense, but the idea of AI Automations that could:
- Scrape a Twitter feed
- Pass structured data into AI
- Generate an essay draft
- Save it in a document
- With one click, in under 2 minutes, from a browser extension
...in 2023, this was insane!

Bardeen was surely ahead when it came to innovating. 

Great times!

Early ChatGPT Innovations

Prediction:

The next wave of profitable GTM micro-SaaS won’t be built by developers.

It’ll be built by GTM engineers.

Why?

Because GTM engineers sit at the intersection of:
→ market pain
→ buyer psychology
→ workflow design
→ marketing campaigns
→ automation
→ AI-agent orchestration
→ tooling knowledge (Clay, Relay, Lindy, Exa, Octave, etc.)


Developers build what’s possible.

GTM engineers build what sells.

Totally different game.

And with the tools they already know:
→ Clay
→ N8N
→ Instantly
→ Airtable
→ Zite
→ Lovable
…you no longer need to write 20,000 lines of code.


What you do need is:
→ deep understanding of a niche
→ a workflow that consistently converts
→ clarity on the audience that cares
→ the skill to turn that workflow into a product

Many of the Clay tables and N8N workflows that are converting today will become micro-tools tomorrow.


I've already seen this happen:

→ automation agencies productizing their best flows
→ lead-gen agencies turning workflows into SaaS (like Leadspicker)
→ tools like Trigify emerging directly from real GTM pain

If you’ve ever automated a campaign that worked so well a stranger asked to buy it…

You're already a product person.

You just don’t know it yet.

GTM Engineers should be building SaaS

Just saw OpenAI's attempt to disrupt the AI agent building experience.

Before you believe the 'OpenAI killed Zapier' headlines, let's talk about what they actually shipped.

They released their Agent Kit with 3 pillars:

→ Agent builder

→ ChatKit

→ Eval

Agent Builder is a visual canvas to design how your agents think and act.

Connecting tools, defining logic, and orchestrating workflows. without writing all the glue code yourself.

What they got right:

→ Few agents reach production

→ Current enterprise agent stacks are too complex

→ Evals are critical

What they missed:

→ The problem isn't how fast you can build an agent

→ It's knowing which agents are worth building

→ And creating actually intuitive tools that democratize agent building

"Idea to agent, faster" sounds like a great line from Sam.

But fast ≠ good.

And what they showed isn't even fast, because it's not intuitive.

It might be fast for enterprise developers with the right access and technical knowledge.

But that's exactly the problem.

The real bottleneck for production agents isn't technical speed, it's business logic complexity.

Placing a stopwatch on a demo doesn't help when only developers can understand the demo.

This is built for enterprise developers, not everyday users, with a false promise of simplicity.

Calling something "intuitive" while requiring knowledge of SDKs, Evals, Guardrails, and code?

That's not intuitive.

Intuitive isn't "fewer lines of code."

Actually intuitive means: someone who understands their business problem can build the solution without becoming a developer first.

In their rush to ship faster, they forgot to build for anyone who isn't already a developer.

OpenAI's Agent Kit Review

When all players were done building their text-to-agent interfaces, OpenAI joins the game.

OpenAI is prepping to drop an "agent builder" TODAY at Dev Day.

If true, it's a big deal.

What could change:

→Before: many tools, weekends debugging, friction to manage APIs

→After: One interface, native models, all connectors in one place, click to deploy... directly where the work already happens.

Will it crush N8N/Zapier/Make?

Mmm let's see.

Here's what everyone misses:

These platforms spent 5+ years refining UX, building communities, and  locking in customers. 

They know their users' workflows inside out.

OpenAI shipping an agent builder doesn't mean existing users will immediately flee.

Reality check:

While you're worried about the next gen of AI agents, most businesses are still trying to get basic 2-5 step automations right.

What happens when the agent barriers drop and more people build?

The same thing that always happens:
→ 1000x more agents get built
→ 90% will be garbage
→ 10% will be genuinely useful
→ 1% will be transformational

The winners won't be the ones who built agents first.

They'll be the ones who knew which agents were worth building.

Today might be more important than people realize.

OpenAI's Agent Builder Announcement

Every major automation platform now has an AI copilot.

In 18 months, building workflows went from technical craft to writing prompts.

I've seen it unfold myself → The trend is clear.

🪄 "Vibe automation" is just around the corner, and building automations will be exponentially easier in a few months. 🧵 

2023 → @BardeenAI Magic Box (April)
2023 → @zapier AI Builder (May)
2023 → @make AI Assistant (December)


2024 → @relay Assistant (May)
2025 → @RelevanceAI_ Invent (May)
2025 → @gumloop_ai Gummie Agent (September)
2025 → @clay_gtm Sculptor (September)

Wouldn't be surprised about @n8n releasing this next

🪄 "Vibe automation" is just around the corner, and building automations will be exponentially easier in a few months.

But *knowing what to build* won't get easier, will get harder. 


BEFORE: 

Bottleneck = technical skill

AFTER: 

Bottleneck = strategic clarity.


What does this means for automation builders?

It's time to switch gears...

Less "how to use tools"

More "how to recognize problems"

Because AI can build the workflow.

But it can't tell you if it's worth building.