Why the Data Foundation for Your Agentic Workforce Matters

Matt Fuller, VP, AI/ML Product, Starburst

Generative AI has commanded the spotlight for the last few years, but the next major shift for enterprise technology is already unfolding: the move from analysis to action. The emerging agentic workforce, humans and AI agents working together to reason, decide, and execute, promises to reshape how organizations operate.

To reach that future, companies don’t need more models. They need a stronger data foundation. Data must be governed, contextual, and accessible enough to support AI agents that don’t just generate insights but take meaningful action on them.

Why Now

AI agents are rapidly evolving beyond copilots that assist humans. They’re becoming collaborators that can trigger workflows, resolve issues, and adapt to changing business conditions. This shift introduces a new challenge: ensuring these agents act on accurate, trusted, and context-rich information.

For years, enterprises have invested in analytics systems designed for human interpretation. The agentic era requires something different: data architectures built to support autonomous action. Analysts and data scientists are no longer the bottleneck; the bottleneck is the architecture that makes data accessible, context-aware and governed.

At the same time, the creation of Open Semantic Interchange (OSI) specification is beginning to define how data platforms, catalogs, and business intelligence tools share metadata. The OSI will define how this metadata is exchanged between tools. For AI, this ensures that when different agents and systems exchange information, they interpret it the same way. In practical terms, it’s the bridge that lets agents in different domains understand the same metadata.

For CIOs and data leaders, this is the new frontier: combining robust governance and semantic consistency to enable agents to collaborate safely and intelligently across tools and teams.

Where Most Organizations Stumble

Most enterprise data architectures were never built for this level of autonomy. They depend on centralized data movement, copying information into lakes and warehouses where it can quickly become stale or disconnected from source context.

Agents operating in that environment either can’t find the right data or make decisions based on outdated snapshots. Without a clear, shared understanding of data semantics, one agent’s “customer” may not match another’s, leading to errors that compound as agents act on conflicting definitions.

This misalignment isn’t new. Human teams have long worked around fragmented data sources, competing dashboards, and subtly different interpretations of core business terms. People have compensated with context, double-checking numbers, clarifying assumptions, and reconciling differences through meetings and manual analysis.That tolerance worked when decision cycles were measured in days or weeks. But AI agents operate millions of times faster and at a far greater scale. Without consistent semantics and governance, the same disconnects that humans could reason through manually now propagate instantly and exponentially, turning small inconsistencies into potentially large operational risks.

Simply put, scaling AI without solving for semantics and governance is like deploying a fleet of self-driving cars without giving them the same map.

What CIOs Need to Prioritize

• Adopt a federated data architecture.

Let data remain where it lives, on-premises or in multiple clouds, while providing secure, policy-aware access. This reduces latency, minimizes data duplication, and preserves context.

• Build around data products.

Data products combine governed, high-quality data with business context, rules, metadata, and ownership. They give agents reliable inputs that can be reused across use cases.

• Prepare for Open Semantic Interchange.

Semantic interoperability is the missing link in AI scaling. OSI ensures that business context and meaning travels with the data, allowing agents and applications to understand the data where the metadata may exist in disparate sources. Starburst is a leading and founding participant in this emerging ecosystem, helping enterprises unify access to distributed data while maintaining consistent semantics.

• Embed governance and observability.

Every agent interaction should be logged and explainable. Policies must be enforced at query time, and observability should track lineage, access, and impact to build trust.

• Enable human-plus-agent collaboration.

This isn’t about replacing people. It’s about freeing them from routine decisions. Define new roles to oversee agent performance and refine prompts, rules, and escalation paths.

• Measure the right outcomes.

Go beyond accuracy metrics. Evaluate time-to-decision, cost efficiency, and compliance adherence. These are the measures that show agents are delivering enterprise value.

The Business Impact

When implemented correctly, this kind of architecture changes the speed and confidence of decision-making. Agents can act in near real time because they access live, federated data instead of waiting for centralized refreshes.

For example, Asurion is already seeing this transformation in action. By using a federated data architecture powered by Starburst, their data teams unified access to customer, device, and operational data across multiple environments without centralizing it. This foundation now supports AI-driven agents that proactively identify service issues, recommend next-best actions, and improve response times. What once took hours of manual investigation can now happen in seconds, enabling Asurion to deliver faster resolutions and a more seamless customer experience.

Semantic alignment prevents redundant or contradictory actions, ensuring that every agent works from the same understanding of the data. The result is greater agility and fewer errors, especially in regulated sectors where consistency and traceability are nonnegotiable.

From Queries to Decisions

The goal of the agentic enterprise isn’t just faster queries; it’s better, faster decisions. When data is governed, federated, and semantically interoperable, AI agents can act with the same discipline and accountability as your best-trained employees.

This is how enterprises move from AI experiments to AI operations. Instead of pilots that prove potential, you get systems that deliver measurable outcomes at scale, faster incident resolution, smarter recommendations, and continuous improvement.

Conclusion

The agentic workforce is here. Enterprises that prepare their data foundations today will be the ones leading this transformation tomorrow.

By combining open, federated architectures with semantic interoperability, organizations can empower agents to act safely and intelligently across boundaries. The result is a connected, intelligent enterprise, one where humans and AI work together to turn data into decisions.

Author Bio:

Matt Fuller is the Co-founder and Vice President of AI/ML Products at Starburst, where he leads the product strategy for artificial intelligence initiatives.  With a background in distributed data systems and analytics, Matt helps organizations unlock greater value from their data through practical, AI-powered solutions. He is also the co-author of Trino: The Definitive Guide, published by O’Reilly Media. Prior to co-founding Starburst, Matt held engineering and leadership roles at Teradata, Hadapt, Hewlett-Packard, and Vertica.