AI Agents Need 'Senses' to Succeed, Experts Say

The rush to embrace agentic AI is hitting a major roadblock: a lack of focus on the underlying data needed to make these AI systems effective. Organizations are hyper-focused on the 'brains' (LLMs) and 'bodies' (infrastructure) of AI, while neglecting the crucial 'senses' – the operational data streams that allow AI agents to perceive and navigate their environments. This oversight, experts warn, could lead to expensive failures and even dangerous outcomes.

The Three Critical Senses for AI Agents

According to a new report by Splunk, successful agentic AI deployments hinge on three fundamental sensory capabilities. First, real-time operational awareness is paramount. AI agents require continuous streams of telemetry, logs, events, and metrics across the entire technology stack, not just batch-processed data.

Second, contextual understanding is essential. Raw data streams alone are insufficient; agents must correlate information across domains instantly to separate signal from noise. Finally, historical memory enables agents to understand patterns, baselines, and anomalies over time, distinguishing between routine fluctuations and genuine issues requiring intervention.

Asana's Chief Product Officer, Arnab Bose, echoes this sentiment, stating, "An LLM in isolation is context-starved... It knows how to write, but it doesn't know your business—your goals, your knowledge, your specific approvals, or your historical relationships." Without this contextual grounding, AI models are prone to errors and inconsistent decisions.

The High Cost of Data Debt

The lack of investment in robust data infrastructure is now coming back to haunt many organizations. What was once a minor inconvenience in traditional analytics—slower insights—becomes a major operational risk in agentic environments. "The velocity of agentic AI doesn't hide these data problems; it exposes and amplifies them at machine speed," the Splunk report warns.

Inconsistent decisions, stalled automation, and the need for manual recovery become commonplace when AI agents lack access to consistent, contextualized data. This 'data debt' can lead to agents oscillating between inaction and unnecessary interventions, ultimately undermining the entire purpose of deploying AI in the first place.

The Competitive Advantage: Data-Driven AI

Winning organizations are recognizing that agentic sensing infrastructure is the real competitive differentiator. They are investing in unified data platforms that can handle petabyte-level datasets cost-effectively, with built-in context and correlation capabilities. Traceable lineage and governance are also crucial, providing the ability to answer "why did the agent do that?" for debugging, compliance, and building trust.

Asana's integration with Anthropic's Claude chatbot highlights the importance of providing AI models with the necessary context. By connecting Claude to Asana's "Work Graph," users can create and manage projects through natural language conversations, grounded in the reality of their organization's operations. Human oversight remains a critical component, ensuring that AI actions are reviewed and approved before implementation.

Furthermore, Asana is not limiting itself to a single AI provider, offering integrations with ChatGPT and Google Gemini to avoid vendor lock-in. "Our philosophy is to meet users where they want to work," says Bose. This multi-provider approach underscores the growing recognition that the value lies not just in the AI models themselves, but in the data and workflows that make them useful.

The key question moving forward isn't just how many AI agents an organization can deploy, but whether those agents have the necessary 'senses' to accurately perceive and understand their environment. Without this foundational capability, the promise of agentic AI will remain unfulfilled, replaced by costly and potentially chaotic outcomes. Companies that prioritize operational data as critical infrastructure will be best positioned to leverage AI effectively, reliably, and at scale. The competitive landscape of 2026 and beyond will be defined not by AI sophistication alone, but by the quality of data fueling these intelligent systems.