Synopsis: In this Techstrong.ai Leadership Insights interview, pgEdge CEO Phillip Merrick explains how the open source Postgres database has become a de facto standard for deploying AI applications. He discusses Postgres’ flexibility, ecosystem momentum, and its growing role in supporting distributed data architectures required for modern AI workloads.
While early AI conversations centered on vector databases and specialized data stores, Merrick argues that Postgres already had the momentum before generative AI took off. Developers were standardizing on it for its flexibility and extensibility, and that foundation carried forward into AI workloads. Rather than being displaced, Postgres has expanded to absorb new requirements through extensions that support vector embeddings, document models and distributed architectures.
Merrick also highlights a more structural shift: AI applications introduce non-human users. Agents continuously query, act and interact with systems in ways traditional applications did not anticipate. That changes how teams must think about authentication, governance, performance and data access patterns. Managing agent identities and enforcing least privilege access becomes as important as schema design.
At the same time, many AI prototypes are built quickly using cloud-hosted Postgres services, but enterprises still face the challenge of bringing those applications into compliant, production-grade environments. Bridging that gap requires more than database compatibility. It demands alignment with security policies, deployment models and existing infrastructure.
The broader takeaway is that AI is not eliminating relational databases. It is increasing pressure to make data accessible in standardized, governed ways. Whether through Postgres or another unifying layer, organizations must rethink how siloed systems expose data to models and agents.
As AI moves from experimentation to production, the database is once again at the center of the conversation.
