Businesses Race to Adopt Agentic AI, but Data Governance Lags

Agentic AI is rapidly becoming a focal point in enterprise technology, with many businesses eager to integrate these advanced systems. According to a survey conducted by PwC in 2025, 79% of senior executives reported that their organizations are already adopting AI agents. However, as experts from AtScale highlight, many companies are moving forward without the necessary data governance frameworks in place, which could jeopardize the effectiveness of these systems.

Understanding Agentic AI

Agentic AI encompasses artificial intelligence systems capable of making decisions and taking actions autonomously to achieve specific goals, without the need for constant human oversight. Unlike traditional AI, which typically responds to prompts, agentic systems can execute multi-step processes and adapt their strategies based on varying conditions.

The challenge lies in the fact that when AI agents operate without proper semantic context, they frequently act on incomplete or misinterpreted data. For instance, an autonomous agent might confidently execute a marketing campaign based on customer segments it has fundamentally misunderstood. This can lead to significant errors and inefficient use of resources.

Companies like AtScale are addressing these issues by creating a semantic foundation that organizations need before deploying autonomous AI. Their platform establishes a universal layer of business context, ensuring that AI agents comprehend not only the data they access but also its relevance to the business.

Operating Dynamics of Agentic AI

Traditional AI systems function much like advanced calculators, responding to specific queries with outputs derived from their training data. For example, a chatbot might provide market analysis, but the user must decide how to use that information. In contrast, agentic AI systems begin with a high-level goal and independently develop a strategy to achieve it through planning and execution.

The architecture of agentic AI involves three core components: planning, execution, and feedback loops. The planning layer breaks down complex objectives into manageable tasks, while the execution engine carries out those tasks in conjunction with various tools and databases. Feedback loops continuously monitor results and adjust strategies in real-time. Despite this autonomy, the reliability of agentic systems hinges on the quality of their data and semantic understanding.

Dave Mariani, co-founder and CTO of AtScale, emphasizes that creating effective AI agents for enterprises is a complex endeavor. He states, “It requires accuracy, governance, lineage, and security. It also necessitates a bridge between natural language and business logic without compromising on standards.”

According to Gartner, agentic AI is currently in the experimental phase of its maturity roadmap. Most organizations remain at Level 2 (AI Assistants), gradually testing Level 3 capabilities, where agents can reason through undefined tasks and collaborate across systems. Gartner anticipates that by 2028, 15% of daily work decisions could be made autonomously through agentic AI, a significant increase from virtually none in 2024.

Early adopters of agentic AI are focusing on three main applications: automated reporting systems that create financial summaries and performance dashboards without human involvement; supply chain monitoring agents that track inventory levels and shipping delays; and advanced customer support routing, whereby agents categorize requests and direct them to the appropriate specialists.

Despite the promising results highlighted in PwC’s survey—where two-thirds of organizations using AI agents reported measurable productivity gains—the success of these implementations heavily depends on data quality and contextual accuracy. Gartner warns that over 40% of agentic AI projects may be canceled by the end of 2027 due to escalating costs, unclear business value, or inadequate risk controls.

The Importance of Semantic Context

One of the critical challenges with agentic AI is the lack of proper semantic context. AI agents that operate on raw data often encounter conflicting metric definitions across different departments. For example, one AI agent might recommend increasing the marketing budget based on favorable customer acquisition costs, while another suggests cutting it due to perceived high costs. This contradiction arises from differing calculations—marketing may consider only direct ad spend, while finance includes salaries and overhead.

When AI agents lack semantic guardrails, they risk making misguided decisions based on varying interpretations of the same metrics. Research indicates that enterprise AI systems achieve only 16% accuracy when querying raw data, compared to 54% accuracy when utilizing structured knowledge graphs.

Semantic layers help mitigate these issues by establishing a shared business language. This ensures that AI agents have a clear understanding of what terms like “revenue,” “customer,” and “conversion rate” signify within the organizational context. AtScale’s semantic layer platform is designed with this principle in mind, empowering AI agents to work with consistent definitions rather than competing interpretations.

Mariani notes, “What the industry calls ‘agentic AI’ today is ultimately an evolution of something we’ve always believed: analytics systems should be intelligent, explainable, and grounded in business logic. The key to that? Semantics.”

Benefits and Challenges of Agentic AI

When semantic layers and agentic AI are integrated, organizations can unlock capabilities that neither technology can provide alone. This combination enhances trust in AI decisions, as it offers transparent and traceable logic behind recommendations. Additionally, teams can experience early efficiency gains, reducing time spent on manual reporting and analysis tasks.

However, the journey to successful agentic AI implementation is fraught with challenges. Companies often discover that autonomous agents introduce complexities that traditional IT governance frameworks cannot adequately address. Governance becomes increasingly complicated when agents operate across multiple systems, necessitating the translation of corporate policies into machine-readable rules.

Moreover, data quality issues can escalate rapidly at machine speed. An autonomous agent may execute campaigns based on flawed segmentation data before any human analyst can identify the problem. Security concerns also arise, as agents require sufficient access to function effectively, but this access must be carefully managed to prevent potential vulnerabilities.

As organizations implement agentic AI, explainability becomes paramount. Stakeholders must understand the rationale behind recommendations that impact significant business outcomes. Without semantic foundations, these decisions can become opaque, leading to a decline in organizational trust.

The rise of agentic AI presents significant opportunities, but it also requires careful planning and implementation. Organizations that have successfully harnessed agentic AI have prioritized establishing a solid semantic context, ensuring consistent definitions and data governance across all systems before allowing AI agents to operate independently.

For companies eager to explore the potential of agentic AI responsibly, establishing a governed semantic foundation is crucial. AtScale’s platform offers the necessary infrastructure to transform pilot programs into scalable, production-ready solutions that work effectively across enterprises.