Context Is The Competitive Advantage AI Models Can't Buy

Sundar Balasubramanian, senior vice president of product management at Icertis, makes the case in a Forbes Technology Council column that enterprise AI value is leaving the model layer. Bigger models and more agents are not the differentiator anymore, he wrote. Foundation models are becoming commoditized. What separates winners now is context: the connected operational data that tells an AI what matters in a particular business, not just what is.

The years of racing to adopt AI gave way to pressure for proof. Boards are demanding returns, Balasubramanian wrote, and CEOs are pushing teams beyond pilots. He rejects the premise that the answer lies in a bigger model or a newer platform. Context, he argues, is not data in the abstract; it is "the full operational reality of the enterprise." Most companies have built the data infrastructure, he writes; what they lack is a way to make the knowledge inside it usable by AI.

Balasubramanian breaks context into three layers. The first is governance: the rules of the business, including how work is structured and how obligations are enforced, much of it codified in contracts. Applied to AI, those rules become guardrails, shaping outputs and keeping them aligned with how the business runs today. The layer extends to business workflows and approval processes, plus the security controls that enforce compliance and accountability, he wrote. Governance is not static; it shifts when companies enter new markets or restructure relationships.

The second layer is decisions. Every organization carries a history of what it accepted and where it drew lines. Those terms and exceptions, often buried in approvals and redlines, tell an AI what this company would actually do. Recommendations stop feeling generic; they reflect the company's own logic instead of a plausible answer. AI without the decision layer can generate answers; it cannot replicate judgment, he wrote.

The column's third layer is patterns, the one Balasubramanian calls most overlooked. He gives examples: repeated deviations from standard contract terms, drift from internal playbooks, regional differences in discounting, delays in a workflow stage such as billing. Patterns surface what nobody documented. A policy can say one thing while the pattern of behavior says another, he wrote; the gap between intention and reality is where risk hides. Patterns, once visible, turn isolated decisions into a view of how the whole business runs.