← all writing
artigo07 Aug 2025

The Semantic Convergence: Why 2000's Metadata is 2025's AI Gold

How SBOM, Schema.org, and controlled vocabularies from the past are the missing key to AI's intention problem

IA:Plano de Adoção - Práticas

AI Governance - Plano de Adoção - Conteúdo voltado para Adoção frente ao desafios de Governança para IA

How SBOM, Schema.org, and controlled vocabularies from the past are the missing key to AI's intention problem


Everyone's talking about AI agents, automation, and the "death of RPA." But they're missing the elephant in the room.

The solution to AI's intention capture problem isn't in the future - it's in the metadata standards we built 20 years ago.

The Hidden Connection

Let me connect some dots that seem unrelated:

2000: Opalis (→ MSOrchestrator)

  • What we built: Workflow automation with structured data

  • What we learned: Metadata matters more than logic

vista

2011: Schema.org Launches

  • What we built: Universal vocabulary for the web

  • What we learned: Shared semantics enable understanding

2018: SBOM Becomes Critical

  • What we built: Software transparency through structured manifests

  • What we learned: Controlled vocabularies prevent chaos

2025: AI Can't Capture Intent

  • What we're missing: All of the above

  • What we need: Semantic structure, not more parameters

"The New Bluescreen of Death"😄 Only those who lives it know

The Convergence Point

Here's what nobody's talking about:

The Uncomfortable Truth

Automation didn't die. It evolved.

  • 2000: Opalis orchestrated systems

  • 2010: Schema.org orchestrated meaning

  • 2020: SBOM orchestrated dependencies

  • 2025: We need to orchestrate intention

Why This Matters Now

The most valuable asset isn't:

  • ❌ More training data

  • ❌ Bigger models

  • ❌ Faster inference

It's:

  • Structured Semantic Layers

Think about it:

  1. SBOM gives us dependency graphs → AI needs relationship graphs

  2. Schema.org gives us shared vocabulary → AI needs controlled terms

  3. Old orchestrators gave us deterministic flows → AI needs guardrails

The Practical Path Forward

Step 1: Vocabulary First

Before building agents, define your domain vocabulary:

Step 2: Intent Schemas

Structure intentions like SBOM structures dependencies:

Step 3: Semantic Validation

Use 20-year-old patterns for modern problems:

  • XML schemas → LLM output validation

  • SOAP contracts → Agent interaction protocols

  • REST principles → Stateless agent design

The Next Asset

Everyone's betting on:

  • Vertical agents

  • Specialized models

  • Custom training

But the real gold is:

  • Semantic Infrastructure

  • Controlled Vocabularies

  • Intent Schemas

  • Validation Frameworks

Who else sees this convergence?


About the author: 20 years be automation frameworks, bridging technical presales, licensing, security, and now semantic AI. Currently building the vocabulary layer that connects past wisdom to future capability.

[]`s Brito

#SemanticAI #SAGI #SBOM #SchemaOrg #IntentionCapture #AIInfrastructure #FutureOfAI #RAG #KAG