McKinsey article.
This McKinsey article analyzes how organizations can scale agentic AI and capture its full value. While many companies have experimented with AI agents, fewer than 10 percent have achieved meaningful impact at scale. The main constraint is not the technology itself but weak data foundations. As a result, the article argues that scalable agentic AI depends on strong data architectures, governance, and operating models.
Data as the backbone of agentic AI
The article emphasizes that data is the core enabler of agentic AI. Unlike traditional AI systems, agentic AI requires continuous coordination across models, workflows, and data sources. This creates greater demands for consistency, accessibility, and real-time processing.
According to the chart on page 2, 80 percent of organizations identify data limitations as a major barrier to scaling AI. At the same time, fragmented and siloed data systems reduce reliability and increase operational risk. Consequently, organizations must move toward integrated and interoperable data environments that support autonomous decision-making.
Building scalable data architectures
To address these challenges, the article outlines key principles for modern data architecture. These include treating data as a shared enterprise asset, ensuring consistent definitions, and embedding governance directly into systems. In addition, organizations must enable secure access, track data lineage, and maintain auditability across workflows.
A critical component is the development of modular and interoperable systems. These architectures allow AI agents to access and process data efficiently while maintaining control and transparency. Without such foundations, agentic systems may produce inconsistent results or fail to coordinate effectively.
The article also highlights the importance of semantic layers and data products. These tools transform raw data into structured knowledge, enabling agents to interpret information correctly and act reliably at scale.
From pilots to scaled transformation
Scaling agentic AI requires more than technical upgrades. The article identifies four key steps: selecting high-impact workflows, modernizing data infrastructure, ensuring continuous data quality, and redesigning operating models.
Organizations should begin by targeting specific workflows where automation can generate measurable value. As shown in the chart on page 6, AI adoption is concentrated in areas such as marketing, knowledge management, and IT. Focusing on these domains allows companies to validate impact before expanding further.
At the same time, data quality must be continuously monitored rather than periodically corrected. This includes managing both structured and unstructured data, as well as outputs generated by AI systems. Consequently, organizations must adopt real-time validation and governance mechanisms to maintain reliability.
Governance, operating models, and future competitiveness
The article stresses that governance becomes central as agentic AI scales. Organizations must define clear rules for how agents access data, make decisions, and interact with systems. In addition, accountability frameworks are required to ensure transparency and compliance.
Operating models must also evolve. Human roles are shifting from execution to supervision and orchestration of AI-driven processes. This creates hybrid environments where humans and agents collaborate, requiring new skills and organizational structures.
The article concludes that data foundations will increasingly determine competitive advantage. Organizations that invest in strong data systems, governance, and operating models will be better positioned to capture value from agentic AI.
Reference
Tavakoli, A., Goodman, B., Soller, H., Rowshankish, K., Kumar, A., Barreto, C., Parekh, S., & Litta Modignani, T. B. (2026, April 2). Building the foundations for agentic AI at scale. McKinsey & Company. https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/building-the-foundations-for-agentic-ai-at-scale#/