A public reference for operating AI systems with explicit state, governance and auditability.
Operational ontology is not a vocabulary list. It is the control layer that connects Data, Logic and Action so AI can operate inside real organizational constraints.
| Publisher | FSTech |
| Canonical page | https://fstech.digital/ontologia-operacional/ |
| Technical essay | https://fstech.digital/framework/ |
| Status | Public reference, not production code |
| License | See LICENSE |
This repository publishes the public surface of the Operational Ontology Framework:
- the core concepts;
- the D+L+A model, Data, Logic and Action;
- governance artifacts for stateful AI work;
- anti-patterns observed in AI automation projects;
- generic schemas and templates that define the interface of the method.
It intentionally does not publish FSTech production code, client implementations, prompts, agents, adapters, deployment scripts or commercial playbooks.
In short: this repository publishes the map, not the machine.
Most AI projects fail because they automate before the organization has a precise operational structure.
A model can generate text. It cannot infer, safely and consistently, all of the following unless they are made explicit:
- which entities exist;
- what each entity means;
- which rules govern each decision;
- which actions are allowed;
- what must be escalated to a human;
- what changed since the last session;
- what must be written back for the next operator.
The Operational Ontology Framework treats those constraints as first-class artifacts.
Every operational AI system needs three layers working together.
| Layer | Question | Failure mode when absent |
|---|---|---|
| Data | What exists and what does each thing mean? | The model guesses entities and relationships. |
| Logic | What rules, thresholds and routes govern decisions? | The model improvises process. |
| Action | What can be executed, by whom and under which guardrails? | Automation creates uncontrolled side effects. |
A chatbot is not an operational system until Data, Logic and Action are explicit and connected.
The framework uses five governance artifacts.
| Artifact | Role | Volatility |
|---|---|---|
| Pin | Invariants: identity, domain entities, boundaries and non-negotiable rules | Low |
| Spec | Current work: tasks, acceptance criteria, blockers and execution state | High |
| Handoff | Session continuity: decisions, attempts, results and next actions | Append-only |
| Facts | Long-term observed knowledge with source and confidence | Medium |
| Skills | Reusable procedures refined through execution | Medium |
These artifacts are not magic. Their value is operational discipline: every AI action can be traced back to explicit context.
Common failure modes this framework is designed to avoid:
- Prompt as architecture: hiding business rules inside a giant prompt.
- Vector memory as governance: assuming retrieval equals accountability.
- Automation before simplification: scaling a broken process.
- Chatbot theater: conversational UI without operational authority or state.
- No write-back: every session starts from scratch because the system learned nothing durable.
- Unbounded agency: allowing an agent to act without clear escalation and action limits.
docs/ Public explanation of the framework
schemas/ Generic, non-client JSON schema examples
templates/ Minimal public governance artifact templates
LICENSE Usage terms
NOTICE.md Attribution and trademark notice
SECURITY.md Security and disclosure policy
No production implementation is included.
Use it when an AI system must:
- preserve state across sessions;
- operate under domain-specific rules;
- produce auditable decisions;
- coordinate humans and agents;
- execute actions with guardrails;
- improve through write-back instead of prompt drift.
Do not use it for stateless text generation, one-off scripts or workflows where the cost of governance is higher than the risk of failure.
FSTech builds operational AI systems for companies that need more than isolated automations.
Site: https://fstech.digital
Canonical page: https://fstech.digital/ontologia-operacional/
Technical essay: https://fstech.digital/framework/