01 Diagnose
We audit what already exists.
We look at what you have today and which processes live only in your team's heads. No generic questionnaires — we come to understand your system.
Accepting projects · ES · EN · LATAM
Is your AI ready for real customers — or just for the demo?
We build AI systems that decide and act on their own — with a human approving every important step, limited permissions, and a log of everything they do. No shortcuts, no winging it. For teams about to put AI somewhere that matters.
ES · EN — Based in LATAM · 15+ years shipping software
01 · Ladder
Where are you?
Nobody arrives saying “I want an AI system.” They arrive saying “I have this problem.” Find yourself in the phrase — each tier has a clear deliverable, duration, and price.
-
“I have an idea, nothing built yet.”
From: a loose idea. To: clarity on what to build and with what.
Initial assessment — what to build first, a clear build-vs-buy call, and which tools to use.
3–5 days -
“I built something with Cursor or Bolt and I'm scared to put it in front of real customers.”
From: a fragile prototype. To: reviewed code, ready for users.
Full review — code, security holes, architecture, and a clear plan for what to fix first.
1 week -
“I need an AI doing real work, not a demo.”
From: plans on paper. To: a system running with real customers.
Build & launch — an AI system in real customer use, with a document your team can read and maintain.
2–6 weeks -
“It's live and I need it to keep improving.”
From: a live system. To: a system that improves with every use.
Monthly retainer — one fixed point of contact, continuous improvements, and capability expansion.
monthly
02 · Process
How we work — three phases, every project.
— Start small, move fast, scale what works.
02 Structure
We make the unreadable readable.
We turn those processes into clear rules a machine (or a new hire) can follow. The first deliverable is almost always a document anyone can read and review.
03 Launch
We launch with a human in control.
AI systems in real customer use, with a log of every decision, limited permissions, and a preview of the result before applying. No security shortcuts.
03 · Cases
Two portfolio pieces you can audit line by line.
No client logos we can't name. Real code, commit history, changelogs. The evidence lives in the repo.
Open source · MIT · v0.4.0
01 orch-agents
A system that decides which AI handles each task on GitHub and Linear — without the criteria living only in one person's head.
- Problema
- The criteria for which AI handled each task lived only in one person's head.
- Solución
- A text file anyone can read and review, with full change history.
In real customer use · v0.2.0
02 tapioca.money
An AI that moves real money between crypto protocols to optimize yield — with safety brakes.
- Problema
- An AI with access to user money, with no clear brakes on what it could do if something went wrong.
- Solución
- The AI can only move money you authorize, like a credit card with a daily limit — and previews every move before making it real.
04 · Evidence
The numbers that matter to us.
Not impressions per month or “happy clients.” Delivery metrics, auditable.
4
Weeks
Average time from project kickoff to first launch.
18+
AI roles
Roles ready in our open-source project orch-agents — reviewable line by line in the repo.
2
In real customer use
Systems running with real customers, open source code, and public commit history.
15+
Years shipping
We didn't learn AI this month. We've been shipping software to real customers for fifteen years.
05 · What it looks like
The knowledge that lived in one head, now lives in a file.
This is a fragment of the real text file that runs orch-agents. The criteria that used to live only in one person's head — which AI handles each situation — now lives in a file anyone can read and review.
WORKFLOW.md
templates:
tdd-workflow: [coder, tester]
feature-build: [architect, coder, reviewer]
security-audit: [security-architect]
github:
events:
pull_request.opened: github-ops
issues.labeled.bug: tdd-workflow
workflow_run.failure: quick-fix
agents:
routing:
bug: tdd-workflow
feature: feature-build
security: security-audit
default: quick-fixReal excerpt from the repository. Full version at github.com/be-automata/orch-agents.
06 · Differentiators
What makes us different.
01 / 03
We don't lock you into a platform
We don't sell you our own platform. For each project we pick the tools that serve your situation — and we explain why before writing a single line. We serve the problem, not a specific tool.
02 / 03
Security from day one
Limited permissions, preview before applying, human review. Not a checklist at the end — it's the architecture from day one.
03 / 03
Written processes, not in heads
The bottleneck is rarely the AI. It's that your organization hasn't described its own work in a form a machine (or a new hire) can follow. Our first deliverable is almost always a document anyone can read and review.
07 · Contact
Let's talk about your project.
01 Book a call
30 minutes. We assess your case and give you a realistic timeline and architecture.
02 Technical proposal
Within 5 days of the call. Architecture, phases, costs, tools chosen and why.
03 We start
First launch in under 4 weeks from kickoff.
No pitch, no commitment. If we're not the match, we'll tell you who is.