Resources
Material from the systems I use every day, for you to use too. Free, with no sign-up.
Start here
If you only read three, make it these.
10 questions to ask before automating your business with AI
To decide in one afternoon which task to automate first, how risky it is and which pilot to start with.
ES + EN · 582 KB
Small pilots before big projects: how I start with a business
I start with a single task and one number to measure; in two weeks you know whether AI really helps you.
ES + EN · 575 KB
Anatomy of a voice + WhatsApp booking agent
How to build an agent that takes appointments without inventing slots or touching money.
ES + EN · 593 KB
How I work in practice
Seven rules I apply in every system, including my own.
- 01
Context first, in simple files
Before building anything, I write down how your business works in text files that you can open and read yourself. The AI works from that. If you change tools tomorrow, the context goes with you.
- 02
As little AI as possible
Code decides whatever cannot go wrong, such as who the client is or what gets sent and to whom. The AI reads and drafts, which is what it does well.
- 03
I fix the access before the prompt
When the AI gets something wrong, the first thing I check is whether it had the right information at hand. Often the fix is giving it access to that information, and not rewriting its instructions.
- 04
Autonomy is earned one task at a time
Every task starts as a draft: the AI prepares it and a person reviews it before anything goes out. Only when it gets real cases from your business right, checked several times, does it start sending on its own.
- 05
Every mistake becomes a written rule
When something fails, I do not stop at fixing that one case. I write the rule down in a file, and the next job already applies it.
- 06
One number per client, agreed before building
Before we start, we choose a single number from your business, for example appointments booked or enquiries answered. If that number does not move, we both see it.
- 07
One space per client and one person in charge
Each client has their own space, separate from the rest, so their data never mixes with anyone else's. And every system has a named person who answers for it.
Guides (10)
Short PDFs with what I actually apply, in Spanish and English.
10 questions to ask before automating your business with AI
To decide in one afternoon which task to automate first, how risky it is and which pilot to start with.
ES + EN · 582 KB
Small pilots before big projects: how I start with a business
I start with a single task and one number to measure; in two weeks you know whether AI really helps you.
ES + EN · 575 KB
AI agents in production without silent failures
How to know every morning, with evidence, what your agents did and what they left undone.
ES + EN · 595 KB
Privacy checklist for automations that touch customer data
15 checks so your automation uses only the data it needs and nobody sees what they should not.
ES + EN · 593 KB
Your business in plain-text notes: the knowledge base your AI reads
Write down once how your business works and any AI can work with it, today and five years from now.
ES + EN · 601 KB
Work while you sleep: AI agents on a 24/7 server
What a server can get done overnight, and the report that tells you in the morning what happened.
ES + EN · 588 KB
A website live in one afternoon: Next.js and Vercel without surprises
What it really means for a website to be live, and how to check it yourself without relying on whoever builds it.
ES + EN · 577 KB
Anatomy of a voice + WhatsApp booking agent
How to build an agent that takes appointments without inventing slots or touching money.
ES + EN · 593 KB
Claude Code on a server: scheduled jobs and limits
How to leave Claude Code working overnight without it eating the machine or the budget.
ES + EN · 595 KB
The right model for each task: pay for judgment, not for reading
Four model tiers matched to the task, so you spend less and get more right.
ES + EN · 587 KB
Templates and code (6)
Public code so you do not start from zero.
Voice and WhatsApp booking agent (template)
The agent that takes appointments for small businesses, with more than 5,000 tests. Code decides and the model only drafts.
EN
Meeting transcription (template)
Transcribes meetings on your own server, separates who is speaking and leaves the summary in Drive or Telegram.
EN
AI secretary for tutoring (template)
Organises sensitive information and drafts reports; if the privacy check fails, nothing goes out. Synthetic data only.
EN
AI knowledge base (template)
It updates itself every day with official documentation and release notes, and Claude Code reads it through an MCP server.
EN
Claude Code work kit
Subagents on four levels, session save points and other pieces I use every day with Claude Code.
EN
GitHub profile
All my public code, in one place.
EN
Videos (6)
Real systems explained in a few minutes.
YouTube channel
A video every day on how I build AI systems for small businesses.
EN
Too Much AI News? My Claude Code Watcher Reads Only What Matters
A watcher that reads AI news and only tells me what matters.
EN
Five Claude Code Automations a Small Business Can Set Up This Week
Five automations a small business can set up this same week.
EN
Your Business Brain in Plain Text Notes
How to write your business down in plain-text notes that the AI reads and keeps up to date.
EN
I Stopped Paying for Meeting Notes
Meeting transcription and summaries on your own server, with no subscriptions.
EN
Cancel Appointments on WhatsApp, Safely
How an agent cancels appointments on WhatsApp without mistakes: code decides and the model talks.
EN
Tell me about new guides
I write to you when I publish a new guide, and for nothing else.
Would you like something like this running in your business?
Free diagnosis (20 min)