"The management that small businesses need but can't afford to hire."
Every decision-maker — founder, executive, professional — faces the same problem:
They have data. They don't have interpretation.
They look at their financials and see numbers. Not what the numbers are saying. Not which question to ask. Not which decision is available to them right now.
Today, this problem is solved in one of two ways:
- You hire someone — a CFO, a chief of staff, a trusted advisor — who reads the data and tells you what it means.
- You don't have that person — and you decide without interpretation, or don't decide at all.
The gap between those two worlds is enormous. The Polic Framework exists to close it.
For years I worked inside organizations building financial intelligence — dashboards, models, diagnostic systems — for founders who couldn't read their own data. I would sit with people who had every number they needed and still couldn't decide. Not because they were incapable. Because the number and the decision lived in different worlds, and nobody had built the bridge.
I was that bridge. I read the data, understood the context, held multiple interpretations simultaneously, and handed back a decision that made sense. At some point I realized I was solving a structural problem by being a person. That doesn't scale. And it leaves everyone who can't afford to hire that person entirely without the interpretation layer.
When LLMs arrived, I understood that the translation problem was solvable at scale — but only if the system doing the translation had documented knowledge, traceable reasoning, and genuine epistemic accountability. Not just a model that generates. A framework that deliberates. This is that framework.
Polic is an epistemic infrastructure — a methodology for building systems that:
- Organize knowledge with provenance and traceability
- Deliberate across multiple perspectives before responding
- Execute decisions with documented reasoning
- Learn over time through a structured amendment system
The framework has three layers:
Layer 1 — Data
Raw data → cleaned data → management view
(Bronze → Prata → Ouro architecture)
Layer 2 — Interpretation
Data read by multiple perspectives
Diagnosis delivered in decision language
Human in the loop for high-stakes outputs
Layer 3 — Deliberation
Embedded council of curated perspectives
Each with documented DNA, biases, blind spots
Traceable reasoning — not black box
LLMs are the new library.
Libraries solved the problem of knowledge access at scale. LLMs are solving the same problem for interpretation. But libraries needed classification systems to be trustworthy. LLMs need the same — an epistemic substrate that makes outputs traceable, auditable, and reliable.
This framework is that substrate.
Instead of asking a single AI to answer, the framework routes problems through a curated council of perspectives — each with documented knowledge, biases, amplifications, and blind spots. The answer emerges from deliberation, not generation.
Inspired by CDD (Dewey Decimal Classification) and CDU (Universal Decimal Classification), adapted for knowledge organization in LLM-native systems. Each perspective has:
- Primary notation (position in the knowledge map)
- Secondary notation (cross-domain relationships)
- Epistemic nature (perspective / process / field / function)
- Abstraction level (foundational / theoretical / applied / operational)
- Change history (revision vs. paradigm shift)
Every classification decision is traceable:
uri_permanente— persistent identifier across substrateshistorico_notacao— change history with type (revision vs. paradigm shift)episteme_contexto— explicit declaration of non-obvious choicesskos_mapping— W3C SKOS vocabulary for RDF export and interoperability
The framework evolves toward a library that doesn't just store knowledge — it executes. Perspectives are linked to procedures. Procedures have declared human-in-the-loop triggers. The system acts, with humans governing the critical transitions.
As this infrastructure matures, it enables three new professional roles:
Knowledge Administrators
Build and govern the epistemic library
Curate perspectives, validate classifications,
approve amendments
Knowledge Executors
Agents that operate from the library
Execute procedures with documented scope
and explicit human escalation triggers
Knowledge Communicators
Interface between the library and end users
Translate outputs into actionable language
Maintain trust between system and human
The open methodology in this repository is applied commercially through Polic — a service that delivers:
- Dashboard — management view of your business data
- Monthly report — data interpreted in decision language
- Embedded council — deliberation on what to do next
The methodology is free. The implementation is not.
If you want Polic applied to your business: → [contact]
/
├── LICENSE Elastic License 2.0 — free for personal use
├── NOTICE Copyright 2026 Lucas Santos
├── README.md This file — vision and overview
├── METHODOLOGY.md How the deliberative system works
├── ARCHITECTURE.md Technical architecture
└── ROADMAP.md Where this is going
Commercial implementation of this methodology means delivering diagnostics, data analysis, and AI-supported decision support based on your organization's data.
It does not mean — and there is no scenario in which it would mean — access to the deliberative council system:
Not available under any circumstances:
→ Database access of any kind
→ Council member profiles, DNA, or configuration
→ System prompts or council architecture
→ Member content or perspective documentation
→ Internal council infrastructure
What a commercial implementation delivers:
✓ Management dashboards built on your data
✓ Financial and operational diagnostics
✓ AI-supported analysis and interpretation
✓ Decision support grounded in your specific context
✓ Monthly or periodic reporting in decision language
The council is the instrument used to build and validate the methodology. It is not the product. It is not shared. It is not replicated.
Elastic License 2.0. Free for personal use. Commercial use and AI training require explicit permission.
See LICENSE for full terms.
Polic Framework — Lucas Santos — 2026