The conversation around autonomous AI agents has shifted. What sounded futuristic a year ago is now showing up in real business operations. In fact, Capgemini found that 93% of executives believe organizations that successfully scale AI agents in the next 12 months will gain a competitive edge. For Chief Data Officers, this creates both a challenge and an opportunity: how to build a framework that can keep pace with AI ambitions while also scaling trust across the enterprise.

Some teams are just kicking the tires on AI agents. Others already have small pilots in flight. A few are preparing to roll them out at scale. No matter where you are, the same truth applies: Success depends on governance, automation, and access. The hard part isn’t spinning up a pilot. The hard part is creating a foundation that will hold up when you expand to hundreds or thousands of agents running at enterprise scale.

Why pilot success doesn’t guarantee enterprise success 

PwC reports that 79% of organizations are already using AI agents. But here’s the catch: Most of those early projects don’t have the depth of governance, access, or automation needed to sustain long-term success. Pilots can get by with manual workarounds and existing access controls. The trouble starts when the business sees the value and wants more.

At that point, the cracks show. Quality issues that felt like exceptions in a pilot suddenly multiply across the enterprise. Costs go up, business value gets harder to prove, and risks mount. Gartner predicts that over 40% of agentic AI projects will be abandoned by 2027 for these reasons.

The root cause is simple: AI agents don’t behave like people. They don’t stop to ask questions, wait for approvals, or sanity-check results. They consume data continuously, make decisions autonomously, and spread data quality issues at machine speed. That means the guardrails we built for human users — roles, permissions, approval workflows — don’t cut it anymore. 

A framework for scale: Governance, Automation, Access

What’s needed isn’t just “better data management.” It’s a framework built specifically for AI agents, where governance, automation, and access work together like three legs of a tripod.

1. Governance: Machine-readable data contracts

Data contracts — programmatic specifications that define data structure, quality rules, and usage policies — already serve human users effectively. Extending them to be machine-readable enables AI agents to automatically verify that data meets requirements before accessing it. This “governance by design” approach builds trust into every data interaction. 

When agents attempt to access data that violates contract specifications, the system automatically blocks access and alerts responsible teams. This ensures that agents only consume data products that meet established quality and compliance standards.

2. Automation: Automating access through intelligent gateways

While machine-readable contracts define what data agents can access, organizations also need to automate how agents actually connect to that data. A Model Context Protocol (MCP) server is a secure gateway that standardizes and automates the process of exposing governed data and tools to AI agents. It ensures that every AI agent’s access is not only enabled but also secure, contextualized, and fully auditable.

3. Access: Building self-service data ecosystems

Even with governance and automated access in place, agents still need a way to discover what data is available to them. Creating a data discovery marketplace where AI agents can autonomously browse, evaluate, and connect to trusted data products transforms how agents discover and incorporate new data sources.

As business requirements evolve, this self-service approach dramatically reduces the time between identifying a data need and fulfilling it, while ensuring access complies with the organization’s governance policies.

Bringing the framework together

As pilots mature into enterprise initiatives, these three pieces have to work in concert. Contracts define the rules. Gateways enforce them while enabling access. Discovery makes it possible for agents to find the right data fast.

And one important note: prioritize open standards and vendor-neutral solutions. That way, you don’t lock yourself into a proprietary ecosystem that won’t scale with you. The launch of the Open Semantic Interchange (OSI) initiative by Snowflake and Tableau highlights the push for open standards, but it’s important to remember that multiple competing standards will likely emerge, making vendor neutrality even more critical.

Build for scale, start now

Even if you’re not rolling out agents at enterprise scale today, the groundwork matters. The organizations that embed governance, automation, and discovery into their early deployments will be the ones ready to scale seamlessly when the business demands it.

Because here’s the reality: AI agents are coming. The question is whether they’ll amplify your data chaos or operate on a foundation strong enough to scale trust and value across the enterprise.

About the Author:

As Chief Technology Officer at Actian, Emma McGrattan leads the company’s technology strategy, innovation, and product development in support of Actian’s mission to simplify how companies connect, manage, govern, and analyze data to transform businesses. Since joining the company three decades ago, McGrattan has played a pivotal role in the evolution and advancement of its analytics, data integration, and data management solutions, including the Actian Data Platform.

A prominent figure in the database industry, McGrattan is known for her expertise in data architecture, query optimization, and cloud transformation. Her leadership and contributions to these areas are widely recognized, making her a respected voice at technology events.

Passionate about creating a sustainable, inclusive future for technology, McGrattan is a celebrated advocate for women in tech and an active mentor dedicated to fostering inclusive cultures within the industry.