Skip to content

Home / 🟠 AREAS / AI Factory

AI Factory

How Abitello builds its automation: AI agents write the software, the integrations and the playbook pages; a person decides at every gate and approves every release.

Purpose

Abitello's promise is property management that runs largely by itself. The AI Factory is where that automation is made: tenant and owner interfaces, integrations with banks and accounting, document handling, and this playbook. Anything with an end date that is built here is a project under 🔴 PROJECTS.

Processes

Build something with an agent

  1. Trigger: a manual step that repeats, found while describing a process (Process map).
  2. Concept: the agent writes what is to be built and why, as numbered decisions; the owner takes each one.
  3. Plan: the work is split into units of one working session each, every step with its check.
  4. Session: the agent builds one unit and commits it to the repository.
  5. Check: the machine checks before a person looks: tests, or a build of the playbook.
  6. Release: a person approves the push; only then does the change go live.
  7. Result: something live, a commit that records it, and a written hand-over for the next session.

Rules

  • Every repository is in the GitHub organisation of Abitello, and main is what is live.
  • Agents commit, people release. No change goes live without a person's approval at that moment.
  • No secret in a repository or on a page. Secrets live in the password manager (Password security).

Where things live

  • Repositories: the Abitello organisation on GitHub.
  • This playbook: repository ABITELLO-playbook.