Research · 2026–2031
Operational 2031
What should become software once AI can understand and run a company?
AI models can already run a small company. What they can't do is read one: its state, its decisions and who approves what live in an aging ERP, in spreadsheets, on WhatsApp, and in the floor manager's head. Five years to research, in public, how to change that.
- Week
- 1 / 260
- Pieces
- 2
- Experiments
- 0
- Models
- 0
- Live hypotheses
- 1
The question
How do we make an organization readable to a machine, without replacing anything it already has?
How I work
Research
Formalize the primitives of an operation —state, events, goals, constraints, decisions, authority, outcome— and test them against real companies.
Build
Small experiments, one at a time, on real operations. With a baseline and a result, not a demo. Starting with our own clients' systems.
Publish
Every week, what I learned: what worked, what failed and what was missing. In Spanish first. Mistakes included.
- One finished deliverable per week, for 260 weeks.
- Every twelve weeks, a thesis review: continue, narrow, or pivot.
- Nothing is called a success without evidence.
Writing from the research
See all writing →No published pieces yet.
Test the thesis
Do you run an operation?
A plant, a warehouse, a distributor, a construction site. I'd like to test this thesis against your reality. It's not a sales call; it's a one-hour conversation.
Write to me