The article highlights the shift towards context graphs in enterprise software, emphasizing how traditional systems of record—like Salesforce and SAP—are at risk as AI agents evolve. Jamin Ball argues that while these systems won’t disappear, they must adapt to a new standard. The debate centers on the role of decision traces, which capture the nuances of past decisions—exceptions, overrides, and context—currently scattered across Slack threads or personal notes. These decision traces are critical for agents to understand not just rules, but the rationale behind decisions. Context graphs emerge as a solution, providing a structured record of how decisions were made. This living document connects entities and events, allowing companies to audit and learn from past actions. For example, when an agent proposes a discount, it pulls various data points and approvals into a coherent narrative, creating a decision trace. Over time, this builds a searchable library of precedents that can enhance automation and improve decision-making processes. The article argues that established companies struggle to build these context graphs. Their current systems prioritize present data without preserving the context of past decisions. They miss critical connections that span multiple data sources and platforms. New startups focused on agent orchestration sit in the execution path, allowing them to capture decision traces in real-time. The existing players, meanwhile, can only view historical data after the fact, limiting their ability to understand the "why" behind decisions. As a result, the opportunity lies with these new startups to create systems that serve as comprehensive records of decision-making, potentially leading to the next wave of billion-dollar platforms.