After two years of pilots and proofs of concept, in 2026, enterprises are beginning to move agentic systems into production. These agents can reason, plan, and act inside live business workflows. That shift is forcing a hard rethink of how AI infrastructure is designed.
According to EDB's Sovereignty Matters research, 95% of major enterprises want to be their own AI and data platforms within the next 780 days. The ambition is consistent across every global market, from North America to APJ to EMEA.
The good news is that enterprises do not need to invent the data to get there. Yet only 13% have found the effective model so far, and it lives mostly in their transactional data. Of that group, 93% have designed a highly sovereign AI and data infrastructure to deliver: hybrid, secure, and agile. A small selection of that 13% can be found here, and a 15-sec test to see how closely an organization's DNA aligns with the traits for success. The wealth of agentic AI is in the value of an enterprise's transactional data.
Agents Have a Different DNA, Start with Transactional Data
Traditional AI architectures were built primarily to support model training and inference. Agentic systems are different. As Nancy Hensley, chief product officer of EDB, puts it, "Agentic systems are designed to reason, plan, retrieve contextual knowledge, and execute tasks within enterprise environments. Agentic systems interact continuously with operational data, internal knowledge bases, APIs, and business workflows."
That continuous interaction requires infrastructure that can maintain consistency, governance, and traceability across the entire AI lifecycle.
As organizations deploy AI that actively participates in operational processes, the architecture supporting it must evolve into what many now describe as a secure AI and data factory.
Identity, Access, and Data-Level Governance
Modern AI platforms involve many actors, from human operators to data pipelines and agent runtimes, each of which must operate with clearly defined permissions. PostgreSQL's robust role-based access control model helps organizations define roles for users and services and apply least-privilege access, and, combined with row-level security and schema or database separation, enforce tenant isolation.
Beyond identity-based controls, enterprises must also govern access at the data level. PostgreSQL supports this through row-level and column-level access controls, which help organizations limit access to specific records or sensitive attributes such as personally identifiable information, financial data, and regional datasets. These capabilities help organizations comply with privacy, residency, and internal governance requirements.
Security also requires protecting operational credentials and ensuring that the data entering the platform is trustworthy. AI systems interact with sensitive assets such as enterprise datasets, model service credentials, pipeline authentication tokens, and integration keys. PostgreSQL supports secure deployment through TLS and encrypted backups, with encryption at rest available through extensions such as pg_tde, plus integration with enterprise secrets management platforms. A secure AI factory also depends on governed ingestion pipelines that validate raw data before it becomes part of the operational platform, creating a trusted foundation for the models, agents, and workflows that rely on it.
Hybrid Queries an Individual's Eyes and Ears
Modern AI architectures depend heavily on semantic retrieval. PostgreSQL supports this through extensions such as pgvector, which allow embedding vectors to be stored and queried alongside relational metadata.
This enables hybrid queries that combine semantic similarity search with traditional SQL filtering. An AI system might retrieve documents semantically related to a query while also ensuring those documents meet specific security or regulatory criteria. Integrating vector search directly within the database simplifies the architecture while ensuring governance policies remain enforced during retrieval. Hybrid queries will dominate the future, positioned as a platform's eyes and ears.
Traceability That Can Be Reconstructed
"Enterprises must be able to reconstruct how a particular output was generated, which data sources were used, and which models or prompts were involved," explains Priyanka Jain, EDB's VP of Product, Data & AI Governance.
PostgreSQL can serve as the system of record for this traceability, storing lineage metadata that captures dataset versions, prompt revisions, model versions, training parameters, and evaluation outcomes. By linking these artifacts together, organizations can recreate the chain of events behind a particular agent output.
Beyond lineage, organizations must maintain comprehensive operational audit logs for their agentic systems. PostgreSQL supports audit logging through database auditing extensions and append-only event tables that capture system events and user actions, data access, model invocation, agent execution, and configuration changes. Append-only, tamper-evident logs, enforced through table permissions and triggers, help organizations investigate security incidents and demonstrate compliance with governance policies.
Orchestrating Agents at Scale
As AI platforms grow more complex, the orchestration of pipelines, models, and agents becomes increasingly important. PostgreSQL can serve as a central orchestration registry for job queues, execution dependencies, and pipeline run manifests. This helps teams track progress, coordinate workflows, recover from failures, and reduce the risk of "agent sprawl" as fleets of agents multiply across the business.
Sovereign by Design
Finally, enterprise AI infrastructure must support resilient and sovereign deployment models. Many organizations operate across multiple geographic regions and must comply with regulations governing data residency and processing. PostgreSQL supports these requirements through streaming replication and point-in-time recovery, with automated failover provided by enterprise offerings, enabling organizations to deploy AI platforms across distributed environments while ensuring high availability and disaster recovery. Combined with enterprise platforms such as EDB Postgres® AI, this can provide a globally distributed infrastructure that maintains operational continuity and regulatory compliance.
An Agentic Future: The Database As Center of Gravity
The rise of agentic AI is redefining the role of the database within enterprise architecture. Databases are no longer simply systems of record; they are becoming the coordination layer that enables autonomous systems to operate safely within complex business environments. PostgreSQL's extensible architecture allows it to support transactional data, semantic retrieval, governance controls, and operational workflows within a single platform.
For organizations seeking to build secure, sovereign AI factories capable of supporting enterprise-scale autonomous systems, the database has become the center of gravity for the entire AI infrastructure. The factory is ready to be built, and the blueprint already exists within the transactional data.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
