Presentation: Architecting the Data Layer for AI Agents: From Transactional Systems to MCP and Semantic Models
Enterprise data layers aren't ready for agents. Here's how TOTVS is rearchitecting to feed token-hungry autonomous systems without breaking the bank.
Why it matters
As agents move from pilot to production, the data infrastructure underneath them becomes the bottleneck. This talk walks through real patterns for context optimization, token efficiency, and safe agent access to live transactional systems — the unsexy but critical layer most practitioners ignore.
The key facts
13 to knowData mesh architecture for agent-native data access
Low-latency database optimization for real-time agent decisions
Semantic ontologies and dynamic MCP tool selection for context window efficiency
Balancing deterministic logic with non-deterministic LLM behavior
Token overhead reduction in transactional systems
Security and precision constraints in enterprise agent deployments
Speaker: Fabiane Nardon (TOTVS)
Published: August 29, 2026
TOTVS case study on enterprise data preparation for agents
Data mesh and low-latency database architectures for agent context
Semantic ontologies and dynamic MCP tool selection
Token optimization and context-window management
Focus on precision, security, and cost in agent deployments
Go to the source
InfoQ AI/MLinfoq.com
Publisher excerpt: Fabiane Nardon shares how TOTVS prepares enterprise data for token-hungry AI agents. She discusses balancing deterministic logic and non-deterministic LLMs across precision, security, and cost. Nardon details using data mesh, low-latency database architectures, semantic ontologies, and dynamic MCP…