AI is Fundamentally Changing the Face of Data Management
As AI Workloads Reshape Infrastructure, AI Agents Reshape Data Operations
The Long & Short
While AI workloads cause dramatic shifts in database needs, AI-powered capabilities are coming online to help get the job done. Organizations must invest now in both sides of this coin to keep pace with the competition and future-proof their database operations in the face of an uncertain road ahead. Mature, open source solutions such as PostgreSQL (+ pgvector) offer the performance, flexibility, and reliability organizations need to meet this mark. At the same time, increasing the use of AI-powered automation across one’s database management operations has become critical for overall agility and rate of innovation.
Global spending on data infrastructure has nearly doubled over the past five years (from ~$180 billion in 2019 to $350 billion in 2024), due in large part to AI and analytics needs. Looking ahead, McKinsey estimates that 70% of all data center capacity demand by 2030 will be to host AI workloads, and demand for AI-ready capacity could grow ~33% annually through 2023–2030. It should come as no surprise that this dramatic growth is having profound effects on the scale and nature of data infrastructure.
As organizations race to develop external AI offerings and internal capabilities alike, the need to accommodate AI workloads has become paramount for database management systems and the tools and services that surround them. At the same time, AI-powered tools and capabilities are being developed at breakneck speed to help database administrators (DBAs), developers, and other end-users automate and accelerate the management and optimization of their database deployments.
How AI Workloads are Changing the Face of Enterprise Database Architectures
Artificial Intelligence (AI) workloads are reshaping the expectations placed on enterprise database architectures. Unlike traditional analytics pipelines that operate on static, structured data, AI workflows demand dynamic, distributed, and scalable access to a broad variety of data types — from unstructured documents and time-series telemetry to embeddings and vectorized representations of text, image, or behavioral data.
The most successful AI initiatives are tightly coupled with the organization’s ability to rapidly collect, transform, store, and retrieve data — often in real time. New applications, particularly in areas such as customer intelligence, generative AI, and supply chain optimization, depend on:
- High-throughput ingestion and low-latency reads/writes
- Support for vector similarity search
- Scalable metadata and feature stores
- Seamless integration with machine learning pipelines and model APIs
These evolving requirements necessitate a fundamental rethinking of enterprise database architecture. And the market is responding. New, proprietary vector databases and other, AI-centric solutions are flooding the market as entrepreneurs seek to capitalize on this sudden shift in infrastructural needs. However, it’s the established, open source solutions that appear to be winning out.
PostgreSQL, for example, with its unsurpassed extensibility and rich ecosystem—most notably its pgvector extension—has emerged as an early favorite in the race for AI workload dominance. In a study recently conducted by Percona, 78% of respondents said that PostgreSQL was important to their organizations’ current or future AI/ML initiatives, with 25% saying it was “mission critical”.
In a field as nascent and fast-moving as AI, organizations are rightfully weary of building their operations on the back of unestablished, proprietary solutions that risk things like vendor lock-in, fail-states, and lack of available talent in the labor pool.
As a result, open source solutions like PostgreSQL are winning the battle for AI workloads. With their large, well-established communities, and extreme adaptability, these solutions offer a degree of flexibility and future-proofing that simply wouldn’t be possible with proprietary solutions. Moving forward, we’ll surely see more entrants and upheaval in the field, but if these early figures are any indication, it would seem that open source will indeed be foundational to AI development for many years to come.
How AI-Powered Capabilities are Redefining Database Management
At the same time that AI workloads are reorienting organizations’ database management needs, they’re also redefining the capabilities of database management solutions themselves. While generative AI is already helping to democratize database management through the use of natural language processing and prompting, the future of AI-driven database transformation undoubtedly belongs to agentic AI.
Though still in its early stages, agentic AI is already showing transformative potential in database management, particularly in environments where uptime, performance, and scalability are non-negotiable. Emerging capabilities include:
- Autonomous workload tuning based on usage patterns
- Anomaly detection and failure prevention
- Dynamic query and index optimization
- Automatic response to alerts and remediation
While these advances hint at self-healing, self-optimizing databases, successful implementation will require careful oversight. Human involvement remains essential — not just for safety, but also for ensuring interpretability and accountability. The open source community is poised to play a pivotal role in this process.
Just as it has shaped the generative AI landscape, open source will be vital in creating transparent, ethical, and adaptable AI systems. It fosters innovation, prevents vendor lock-in, and serves as a safeguard against the opacity of black-box models. By enabling organizations to audit, modify, and contribute to shared tools, open source encourages AI democratization while helping mitigate the risks of autonomous systems.
What Organizations Can Do to Keep Ahead of the Curve
AI is already having a tremendous impact on both enterprise database architectures, and the way that teams manage and interact with those systems. In both the near and long term, organizations that adopt the right tools and approaches around AI will gain potentially unparalleled competitive advantages. Regardless of size, industry, or strategic initiatives, there are certain things every organization can do to gain that needed edge, including:
- Unify and Streamline Your AI-Adjacent Tooling, Infrastructure, and Services: In an open source, multi-database environment, choosing the right observability, monitoring, and management solutions—especially those that can work to unite and streamline your operations across DBMSs—will be critical to limiting TCO, breaking down data silos, and accelerating the pace of innovation.
- Lean on Established, Open, and Extensible Solutions: While no one solution will be enough for all of your AI needs, certain core DBMSs stand out as critical for long-term AI stability and success. As we’ve seen above, PostgreSQL has already proven itself indispensable for many organizations’ AI workloads. With extensions like pgVector for embedding similarity search, PostgreSQL can store and query AI vectors, enabling use cases like semantic search. Developers building AI agents often choose PostgreSQL as the de facto storage and search DB, taking advantage of its integrations with AI tools like LlamaIndex and LangChain.
- Look for AI-Enabled Tooling to Help Automate More of Your DB Management: As discussed earlier, the integration of AI-powered capabilities into DBMSs and adjacent tooling will soon become mission-critical for organizations’ competitiveness. Democratization of data access through natural language processing and generative AI will empower developers and others to access and interact with databases without the need for already over-stretched DBAs. At the same time, agent-driven automation will empower organizations to do more with less, and dramatically accelerate the SDLC and rate of innovation.
- Use AI to Reinforce, not Replace, Human Talent: While AI agents continue to mature and develop, being too hasty or overzealous in the adoption and implementation of AI-powered automation poses serious risk. These still largely untested systems require human oversight and should be treated as a means of augmenting and supplementing human expertise.
The potential impact of AI on the world of database management is hard to overstate. Over the coming months and years, we will undoubtedly see more and more agent-driven automation introduced into the space. And while this undoubtedly holds great promise, it is essential that organizations proceed with pragmatism. Rigorous testing and validation alongside human oversight and control will remain critical for the foreseeable future. Ultimately, the mindset behind AI use should always be one of human augmentation, not replacement.
Peter Zaitsev
Founder
Percona