The concept isn't about fitting a massive codebase into one context window. Instead, it's a sophisticated architecture using a deep relational knowledge graph to inject only the most relevant, line-level context for a specific task at the exact moment it's needed.

The defensibility of AI-native software will shift from systems of record (what happened) to 'context graphs' that capture the institutional memory of *why* a decision was made. This reasoning, currently lost in human heads or Slack, will become the key competitive advantage for AI agents.

The effectiveness of enterprise AI agents is limited not by data access, but by the absence of context for *why* decisions were made. 'Context graphs' aim to solve this by capturing 'decision traces'—exceptions, precedents, and overrides that currently live in Slack threads and employee's heads, creating a true source of truth for automation.

The LLM itself only creates the opportunity for agentic behavior. The actual business value is unlocked when an agent is given runtime access to high-value data and tools, allowing it to perform actions and complete tasks. Without this runtime context, agents are merely sophisticated Q&A bots querying old data.

The effectiveness of agentic AI in complex domains like IT Ops hinges on "context engineering." This involves strategically selecting the right data (logs, metrics) to feed the LLM, preventing garbage-in-garbage-out, reducing costs, and avoiding hallucinations for precise, reliable answers.

The system ingests a company's knowledge bases to generate an initial "context graph." As the AI operates, it uses LLMs to explore new conversational patterns. Once a pattern becomes frequent, it's codified into the deterministic graph, making the system more efficient and reliable over time.

AI agents are simply 'context and actions.' To prevent hallucination and failure, they must be grounded in rich context. This is best provided by a knowledge graph built from the unique data and metadata collected across a platform, creating a powerful, defensible moat.

Capturing the critical 'why' behind decisions for a context graph cannot be done after the fact by analyzing data. Companies must be directly in the flow of work where decisions are made to build this defensible data layer, giving workflow-native tools a structural advantage over external data aggregators.

Salesforce's Chief AI Scientist explains that a true enterprise agent comprises four key parts: Memory (RAG), a Brain (reasoning engine), Actuators (API calls), and an Interface. A simple LLM is insufficient for enterprise tasks; the surrounding infrastructure provides the real functionality.