Dominic Jainy stands at the forefront of the modern data revolution, bringing years of expertise in artificial intelligence, machine learning, and the burgeoning field of blockchain technology. As an IT strategist, he has navigated the complex transition from static data repositories to the dynamic, autonomous environments that define our current landscape. His deep interest in how these technologies intersect has made him a sought-after voice for organizations looking to bridge the gap between legacy systems and agentic AI. In this conversation, we explore the shift toward agentic data management, a movement that is redefining how enterprises prepare, govern, and utilize their most valuable asset: information.
The following discussion explores the emergence of autonomous agents as the new backbone of data operations, moving beyond simple task automation to full lifecycle orchestration. We delve into the critical “build versus buy” dilemma facing engineering teams, the specialized offerings from major tech players like Google Cloud and Informatica, and the evolving necessity of robust AI governance and FinOps. Dominic also shares insights on how organizations can identify low-risk entry points to build institutional expertise without compromising security or budget.
Agentic systems are shifting from just analyzing data to managing the underlying data lifecycle. What specific operational advantages do you see for organizations that transition to this agentic data management approach?
The transition to agentic data management represents a fundamental shift from human-led data stewardship to a model where AI agents actively maintain the health and flow of the data ecosystem. By deploying these systems, organizations can automate the high-friction chores of data processing, such as mapping complex schemas and classifying sensitive personally identifiable information across massive datasets. According to recent industry observations, including the trends highlighted by Gartner in June, this move streamlines operations by allowing agents to perform the “step back” tasks—managing the very data that fuels other business-facing agents. This creates a self-sustaining cycle where AI readiness is built into the infrastructure rather than being a manual, post-process cleanup. Furthermore, because these agents can monitor data pipelines in real-time and remediate quality issues the moment they arise, the operational speed of the enterprise increases significantly while reducing the risk of downstream errors in analytics or decision-making.
When deciding whether to build custom agents or purchase vendor-supplied solutions, what are the primary maturity indicators an enterprise should evaluate within its own data engineering environment?
The build-versus-buy decision is a nuanced one that rests heavily on how specialized an organization’s workflows truly are. For a company deeply rooted in digital products, there is often a natural inclination to develop custom agents that connect directly to established tools like DBT for data transformation or Apache Airflow for orchestration. However, managing custom code involves a hidden tax, particularly when data governance requirements shift or scale; it requires a level of engineering maturity that can handle the maintenance of bespoke permissions and audit logs. On the other hand, vendor-provided solutions from companies like Acceldata or Informatica are becoming increasingly attractive because they come with built-in lineage capabilities and pre-configured support for regulatory compliance. Most organizations are currently landing on a hybrid approach, where they use vendor-developed agents for routine platform work while reserving custom-built agents for highly specific business rules that provide a competitive edge.
With major players like Google Cloud, Informatica, and Ataccama launching agentic platforms recently, how should data leaders navigate the diverse range of prebuilt agent capabilities now hitting the market?
Navigating the vendor landscape requires a clear understanding of the specific functional “gaps” in your current data stack, as each vendor has carved out a particular area of strength. In April 2026, Google Cloud announced its Agentic Data Cloud, which includes a Data Agent Kit specifically designed for tasks like database provisioning and performance monitoring, making it a strong choice for infrastructure-heavy needs. Meanwhile, Informatica has deployed its Claire Agents, which are specifically tuned for creating data quality rules and building pipelines through a no-code interface, catering to teams that want to move fast without deep coding. Other providers like Ataccama and Alation launched their offerings in 2025, with Alation focusing heavily on the governance of critical data elements for risk management and financial reporting. Leaders should look for “Agent Studios” or development kits that allow for a degree of customization, ensuring that the prebuilt efficiency of a vendor tool can still be tailored to the organization’s unique data signatures.
As agentic data management moves from pilot programs into full production, what are the most significant governance and operational hurdles that teams are encountering?
The biggest hurdle is not the capability of the AI models themselves, which are already quite advanced, but the operational structure—or lack thereof—surrounding them. As Sami Rahman from DataArt noted, organizations are realizing that they need reliable evaluation frameworks and very clearly defined escalation procedures for when an agent encounters an anomaly it cannot solve. We are currently in a phase where agentic data management is serving as a critical proving ground for enterprise AI governance, forcing companies to address how they monitor autonomous actions. There is also the significant challenge of AI FinOps; without mature cost management, the computational expense of running multiple autonomous agents can spiral out of control if not properly bounded. Additionally, because most current agents are designed for single, siloed tasks, the industry is still waiting for the full maturity of multi-agent workflows that can collaborate seamlessly across different parts of the data lifecycle.
For a data team looking to start their journey with agentic AI, how do you recommend they select their initial use cases to ensure they build expertise without taking on excessive risk?
The most successful strategy for early adoption is to focus on agents with a high anticipated value but where the potential costs and risks are naturally bounded. Data teams should look for well-defined, repetitive tasks such as identifying data quality issues, schema mapping, or basic pipeline monitoring, as these provide a solid foundation without giving the agent full, unsupervised control over mission-critical systems. In practice, this looks like a “human-in-the-loop” model where an agent completes the first pass of a task—perhaps drafting a data transformation rule or profiling a source—and then an experienced data engineer reviews and approves the output. This approach allows the team to gain confidence in the technology while the agent handles the heavy lifting of data preparation that happens before AI model training. By starting with these foundation-layer use cases, organizations can gradually scale to more complex, multi-agent optimizations as their internal expertise and governance frameworks catch up.
Looking at the current landscape and the rapid evolution of these tools since 2025, what is your forecast for how agentic data management will transform the role of the data engineer over the next few years?
I expect that within the next two years, agentic capabilities will become a standard, almost invisible functionality within every major data platform, effectively shifting the data engineer’s role from a “builder of pipelines” to an “architect of agents.” We will see a move away from manual coding for routine tasks and toward the use of graph analytics and model routing to manage complex, collaborative agent ecosystems. The primary point of differentiation for companies won’t be whether they use AI agents, but the quality and rigor of the governance controls they build around those agents to ensure transparency and reliability. Eventually, the manual review of every agent-generated output will fade into exception-based management, where engineers only intervene when the system’s confidence scores drop below a certain threshold. This evolution will allow data teams to manage exponentially larger volumes of data with the same headcount, finally closing the gap between the massive influx of raw information and the actionable insights businesses crave.