DEMOCRATIZED DATA ANALYTICS
Agentic AIs are democratizing data analysis, said Jeff Foster, director of technology and innovation at Redgate. “This is brilliant and empowered citizen developers can often build sophisticated tooling to solve their own problems. Agentic solutions seek to address the growing issue of database sprawl that plagues enterprise data estates.”
Why it matters—Databases are growing too large and complex for DBAs to reasonably manage, Foster explained. “AI cuts through the noise at-scale, supporting monitoring efforts by parsing out which alerts are high-priority, and which can be ignored. Looking beyond, agentic solutions begin to close the gap between developers and DBAs. Developers often struggle to understand the nuances of the database. AI can act as a safeguard, catching problems before they spiral out of control.”
Potential issues—From a data management perspective, AI agent-driven democratization spurs challenges “around how data is being used, where data is being used, and increased demand from professionals across the organization,” said Foster. Then, there is peoples’ willingness “to give up control to a robot. Granting autonomous access to sensitive databases is a high stakes leap that, even if the technology is solid, many leaders are not ready to make.”
Advice—If your current processes don’t work for humans, adding an agent is unlikely to help, Foster cautioned. “The most common roadblock to success I’ve observed is a fragmented estate without consistent processes. If DBAs can’t keep track of data or keep it clean, AI will only make existing problems worse.”
OBJECT-BASED STORAGE FOR AI
Object-based storage may be the key to fueling inference driven AI pipelines. Witness the shift of NVIDIA’s reference architectures “from file to object storage, with NVIDIA beginning to certify object storage for GPU [graphics processing unit] direct access,” related Paul Speciale, chief technology evangelist at Scality.
Why it matters—Today’s AI workloads—particularly retrieval- augmented generation—“depend on fast, reliable access to vast volumes of unstructured data,” Speciale said. This is especially critical to supporting LLMs that need to underpin large userbases. In this case, “storage demands to retain and manage conversational state will quickly exceed GPU memory and even multi-petabyte all-flash systems.”
Potential issues—“Many enterprises face legacy infrastructure constraints, integration complexity, and the need to rethink data life cycle management,” Speciale cautioned. Object storage implementations also require specialized skills.
Advice—“This shift reflects a broader realization, that AI success is less about models and more about data readiness at scale,” said Speciale.
SEMANTIC MACHINE LEARNING FOR DATA SECURITY
The integration of localized, semantic machine learning (ML) into data security is an emerging development, said Stanislav Kazanov, head of GRC, cybersecurity, and sustainability at Innowise. “We have finally moved beyond basic regex searches—such as looking for a 16-digit credit card string—to semantic data discovery.”
Why it matters—Enterprise businesses are overwhelmed with dark data “that is made up of petabytes of unstructured, undocumented data stored in legacy SharePoint drives, Slack channels, and [Amazon] S3 buckets,” Kazanov explained. “Through the use of context-aware ‘data security posture management,’ organizations are able to automate the scanning of these unstructured repositories for the purpose of understanding the context.” For example, he illustrated, “the AI will recognize that a paragraph about a medical diagnosis that appears directly adjacent to a person’s first name is highly regulated protected health information, regardless of whether or not there are any explicit tags.”
Potential issues—“Companies face a significant investment in infrastructure as they must have enough computational power to run the semantic classification models locally or within strictly separately housed, private environments,” said Kazanov.
Advice—“First, completely sanitize your data governance before [being] able to confidently employ agentic AI to assess your environments,” said Kazanov.
ORCHESTRATION OR ‘HARNESSES’
Orchestration—“whether it’s called a ‘harness’ or another term du jour”—is key to managing data in an AI world, said Mayada Gonimah, co-founder and CTO of ThreadAI and former head of AI and ML engineering at Palantir. “Orchestration platforms create the infrastructure necessary for AI agents to act independently within defined trusted boundaries. This controlled autonomy relies on three pillars—control, governance, and reliability.”
Why it matters—The rapid proliferation of AI agents “has created sprawling workflows based on scattered data,” Gonimah said. “That disconnected data is difficult to manage and analyze, increasing security and operational risk and undermining the ultimate goal of autonomy.” Orchestration assures control, governance, and reliability, she continued.
Potential issues—As with all changes, “there are potential hurdles to prepare for,” Gonimah cautioned. “The technological challenges mirror other AI programs in that they require quality data, security by design, and quality pilots.”
Advice—“On a behavioral level, humans have to transition from being above a system to part of it,” Gonimah advised. “Rather than reviewing discrete outputs, individuals will need to be actively involved in drawing the boundaries of each workflow.”
BLOCKCHAIN-BASED DATA INTEGRITY
Blockchain may no longer be at the forefront of the industry’s attention, but it is evolving behind the scenes, promising extended capabilities for data environments, especially for data integrity. “Blockchain-based data integrity solutions allow users to verify that a file or dataset existed in a specific state at a specific time by giving each file a unique, unchangeable cryptographic tag, or seal,” said Thomas Berndorfer, CEO and founder of Connecting Software. “If any part of the file is changed, this seal will never match—meaning that blockchain can provide an assurance that a file is authentic, even when it changes hands.”
Why it matters—Blockchain-assured data integrity “is an important consideration as more companies store their data in third-party SaaS and cloud platforms—when your data is with another company, how can you trust it has remained unchanged?” Berndorfer said. This technology offers “verifiable assurance that data is authentic. When a file is successfully sealed via the blockchain, companies can always verify whether the state of a file has been compromised from the point of completion or approval—allowing companies to avoid costly mistakes and violations.”
Potential issues—“Many businesses still treat internal systems, such as ERPs [enterprise resource plannings] and SaaS platforms, and generated logs as reliable sources of truth,” Berndorfer said.
Advice—“Data verification technologies are a key tool in battling against AI fraud and AI manipulation of data, which will only become more prevalent as AI systems grow more capable. I’m confident that blockchain-based data integrity solutions will become vital to data management in the near future.”