Organizations are increasingly turning to agentic AI to enhance efficiency and streamline operations. This innovative approach employs autonomous agents capable of making decisions, taking actions, and learning independently to achieve specific objectives. Analysts at Gartner project that by 2028, approximately one-third of enterprise software applications will incorporate agentic AI, a significant rise from nearly zero in 2024.
The advantages of agentic AI are enticing. It allows businesses to redirect their focus toward high-impact decisions, accelerate customer responses, and foster innovation and growth. Nonetheless, the transition to agentic AI is fraught with challenges. Gartner warns that more than 40% of agentic AI projects could be scrapped by the end of 2027 due to issues such as hype, cost, and complexity.
Challenges in Implementation
A central issue lies in the necessity for a robust data architecture. The true potential of agentic AI unfolds when multiple agents can communicate and coordinate on a reliable, real-time data infrastructure. Current technology, particularly large language models (LLMs), is not without its flaws. These models can “hallucinate,” generating inaccuracies or misapplying information. Such risks are critical, especially in sectors like healthcare and insurance, where mistakes can lead to significant personal harm or legal repercussions.
Implementing effective multi-agent systems proves to be a formidable task. While many agentic AI projects show promise in prototype stages, they often falter when scaling for real-world applications. For agentic AI to function effectively, it requires access to real-time data from an organization’s on-premises and cloud databases, as well as external data sources. This need highlights a broader transformation occurring in enterprises, where analytics are increasingly converging with operational systems to support agentic AI.
Data Integration and Governance
To fully harness the benefits of agentic AI, companies must unify data from various sources in a trustworthy manner. Establishing strong guardrails, clear permissions, and comprehensive audit trails is essential to ensure data security and accuracy. Organizations need to clean and standardize their data while implementing rigorous data governance practices, enabling them to generate actionable intelligence for task automation.
AI agents cannot function in isolation; they require a shared context, coordinated actions, and real-time decision-making capabilities while integrating with external tools and organization-wide data repositories. Recent open standards in the agentic AI space, such as the Model Context Protocol (MCP) and Agent to Agent communication (A2A), show promise in facilitating effective communication and task execution across complex enterprise workflows.
The rise of a sophisticated AI data layer is necessary, as traditional data platforms like data warehouses and lakes are no longer sufficient. Today’s demands call for data access across various applications, including machine learning, business analysis, and reporting. The primary challenge in constructing data infrastructure for agentic AI lies in operationalizing and scaling it in a cost-efficient manner.
Data governance, access control, observability, and security are crucial elements of this infrastructure. A data fabric architecture may emerge as an optimal strategy for managing this data landscape. This approach ensures that data is consistently accessible and integrated across all systems, eliminating fragmentation and enhancing the overall quality.
Data fabrics utilize metadata management, knowledge graphs, and semantic layers to contextualize data, enabling AI agents to grasp business contexts and interrelationships among various data points. This capability is fundamental for agentic AI systems, which rely on unified data to deliver accurate and contextually aware insights for decision-making and task automation.
Conversely, the data mesh approach offers a decentralized model for data management, where data ownership remains with domain owners responsible for defining and governing their data products. This method relies heavily on people and processes, adding complexity to the coordination required for federated data infrastructures.
Early successes in implementing agentic AI systems have been linked to centralized data infrastructures based on data fabrics. It is evident that effective data governance and ethical frameworks are critical for the safe and responsible operation of AI agents, ensuring alignment with organizational policies and regulatory compliance.
As businesses navigate the complexities of agentic AI, the importance of a reliable and cohesive data strategy cannot be overstated. Organizations that invest in robust data architectures and governance frameworks stand to gain a competitive advantage in an increasingly automated future.
