Context graphs in finance operations
Systems of context will become the most business-critical system in the AI era. A thesis on how this plays out in finance operations.
Systems of context are going to become the most business-critical system in the AI era. Finance operations like order to cash and procure to pay are business areas where context is highly fragmented, and where this thesis applies directly.
Somewhere right now, someone in finance is trying to figure out why a customer is stuck at 62 days overdue. To do it, they're gathering context from the ERP, email chains, external portals, the CRM, maybe a billing system, maybe a conversation with sales. Then they make a judgment call on what should happen next.
Finance teams make thousands of these micro-decisions every week: how to resolve a PO mismatch, who to loop in to chase a customer, when to escalate, what terms to offer. These decisions happen in motion, across systems and conversations. Often, the reasoning behind them is passed on as tacit knowledge, if at all.
In order to cash, no one place captures the full story for a customer:
- Systems: Multiple ERPs, the CRM, billing systems, AR tools, email chains, etc.
- People: Sales, finance, customer success.
- Decision logs: Not captured.
Foundation, Theory Ventures, and others have all written about an emerging thesis: context graphs, business context layers, systems of context.
These are high-context business systems that capture actual decision traces, not just data. The undocumented reasoning that explains how your company actually operates and makes decisions. Inputs, actions, outcomes, and explicit rationale, for every granular decision.
AI agents can now capture these decision traces directly, during task execution. The resulting data set is a queryable context graph. Imagine being able to ask: "Show me all invoices over 60 days where the dispute involved a PO mismatch and sales was looped in. What was the median time to resolution? What are the recurring root causes?"
Once your finance team runs on a system of context, every decision becomes an artefact that both agents and people can learn from. Understanding not just that a process broke down, but why it did. Context graphs also create the loop that allows agents to get smarter with every exception they handle.
The organizations that figure this out will have finance functions that get smarter every week.