Only 10% of Data Teams are Scaling AI – But Pathways are Emerging

Many enterprises are experimenting with agentic or AI-driven approaches to automate workflows and connect data across systems.

Data & AI Strategy Board poll: Where do cross-system data efforts break down most often in your organization? Top answer: Missing/unreliable metadata (lineage, ownership, context), 34%. N=32. Source: Data & AI Strategy Board Community Call, January 22, 2026.

The Story in Brief

“Agentic AI is less a breakthrough solution and more a forcing function that exposes long-standing metadata, ownership, and governance gaps. What stands out is how AI reframes familiar data quality and metadata issues as urgent and unavoidable. Topics that once struggled to gain traction are now top of mind because agents make the consequences of poor foundations immediately visible.”
Sheryl Martin Senior Membership Director, Data & AI Strategy Board

Picking Up the Story

Agentic AI is Stalling at the Enterprise Level

Many enterprises are experimenting with agentic or AI-driven approaches to automate workflows and connect data across systems. But early experiments are revealing how dependent these efforts are on foundational data infrastructure.

45% of members note that their organization is running early pilots that are already hitting foundational limits. Also, only 10% of organizations are actively scaling.

“We see good use cases where the data is actually curated and there’s strong governance, and failures where there wasn’t a clear steward.”

Poll results

Bar chart. Where do cross-system data efforts break down most often in your organization? (select one) Missing/unreliable metadata (lineage, ownership, context) 34%; Org/process friction (ownership, approvals, incentives) 27%; Getting access to data across platforms 13%; Low trust in data quality or outputs 13%; Reconciling inconsistent definitions across domains 13%. N=32. Source: Data & AI Strategy Board Community Call, January 22, 2026.
Cross-System Data Challenges
Donut chart. Which best describes your organization’s current posture toward agentic or AI-driven approaches across enterprise data? (select one) Early pilots — hitting foundational limits 45%; Selective success — works in constrained domains 32%; Mostly conceptual — still defining what’s possible 13%; Scaling — actively expanding across systems 10%. N=31. Source: Data & AI Strategy Board Community Call, January 22, 2026.
Enterprise Readiness for Agentic AI

Focus of the Discussion

AI Agents Are Exposing Your Metadata Problem

Agentic AI is a catalyst that’s exposing weaknesses in enterprise data environments.

Early pilots have shown promise in narrow use cases, but scaling these systems across organizations remains difficult. One of the most consistent challenges involves metadata. Missing lineage, unclear ownership, and incomplete documentation make it difficult for agents to interpret data reliably. As one participant explained, “Context begins to be solutioned via metadata.” Without that context, agents struggle to navigate across systems or produce trustworthy outputs.

Key Takeaways

  1. Gaps in metadata, ownership, and governance existed prior to AI, but were easier to work around in traditional analytics. Agentic use cases often made the consequences of weak foundations more visible and immediate.

  2. Most organizations are finding early agentic pilots work only in narrow, well-curated domains, and stall at enterprise boundaries. Poll results and discussion showed many teams achieving selective success where data is already governed, stewarded, and trusted. Members described a practical “tipping point” where agents function locally, but cannot scale across domains.

  3. Missing or unreliable metadata was cited as the most common point of breakdown in cross-system data efforts. Several members emphasized that master data alone is insufficient without shared semantics that give data meaning across domains. Inconsistent definitions were described as a frequent cause of misleading or unusable agent outputs.

  4. Members pointed to approvals, unclear accountability, and risk posture as limiting factors for agent autonomy. Regulated and risk-averse environments described the need for strict guardrails and human-in-the-loop controls. Many challenges initially framed as “AI limitations” were traced back to process and ownership design.

  5. Several participants shared their belief that a single, universal enterprise agent is unrealistic at this stage. Those participants favored multiple domain- or task-specific agents operating within clearly bounded contexts. Concerns about long-tail scenarios and unintended behavior reinforced skepticism toward fully generalized agents.

About the Data & AI Strategy Board and Assemble

The Data & AI Strategy Board is an Assemble community that helps senior data, analytics, governance, and AI leaders at the world's largest companies make better, faster decisions through trusted peer insights.

We selectively share highlights from confidential member conversations to help leaders understand how peers are navigating complex challenges and fast-moving, high-stakes moments.

If you'd like to learn more about the conversations happening inside our peer network, or how members connect with trusted peers to solve critical challenges, please contact us.

About the author

Jay Avery

Contributing Editor Jay Avery is a freelance journalist and writer whose work has appeared in The New York Times, Vanity Fair, and Architectural Digest. She began her career in marketing and content roles at Uber during its early startup years and later at RideAustin, bringing a strong foundation in both storytelling and strategic communications. Jay holds a law degree from the University of North Carolina and is based in Raleigh, North Carolina.