The HR Context Graph: AI's Missing Layer for People
In the AI era, a context graph captures how decisions get made and why. Here's what an HR context graph is, why you need one, and why it's people ops' missing layer.
Companies are spending billions on enterprise AI—the copilots and agents they’re deploying internally—and getting almost nothing back. MIT’s Project NANDA found that 95% of enterprise AI pilots deliver no measurable return—and the report is blunt about why. The problem isn’t the models, the talent, or the budget. It’s “the lack of learning, integration, and contextual adaptation.” AI is being bolted onto decades-old systems that store outcomes but never the reasoning behind them.
The fix that enterprise software is converging on has a name: the context graph. And while it’s being built for sales, support, and finance, almost no one is building it for people. This guide explains what a context graph is, why it suddenly matters, and what an HR context graph—a context graph for your people—needs to capture to become the missing layer for AI in people ops.
What Is a Context Graph?
A context graph is a living record of how decisions actually get made, stitched across systems and time. Unlike a database that stores the final state, it connects entities—people, tickets, accounts, policies—with the decision traces between them, so AI can see not just what happened, but why it was allowed to happen.
The term was coined by Foundation Capital partners Jaya Gupta and Ashu Garg in a December 2025 essay on AI’s trillion-dollar opportunity. Glean’s engineering team defines it as “a model that connects your enterprise entities (people, documents, tickets, systems) with the temporal traces of actions and events between them.” HubSpot’s Dharmesh Shah put it more simply: “It’s a system of record for decisions, not just data.”
A CRM knows what a deal looks like now. It doesn’t know that you gave a healthcare customer an extra 10% because their procurement cycle is brutal, or that a VP approved the exception on a Slack call. That reasoning—the why—is what a context graph captures.
Why Context Graphs Matter Now
Context graphs matter now because of AI agents. Agents don’t just answer questions, they take actions. To act well without a human in the loop, they need the reasoning behind past decisions, not just the outcomes. Systems of record store the final state; agents need the logic that produced it.
The shift is moving fast, and already faltering. Gartner expects 40% of enterprise apps to feature task-specific AI agents by 2026, up from less than 5% in 2025. But it also predicts over 40% of agentic AI projects will be canceled by 2027, often because the agents lack the context to make good calls.
This is why “context” has become the defining word in enterprise AI. Box CEO Aaron Levie argues we’ve entered “the era of context”: when everyone has access to the same models, the differentiator is the organizational knowledge you can feed them. Forrester analyst Charles Betz calls context graphs “a convergence, not an invention”—the layer agents inevitably need.
What Is an HR Context Graph?
An HR context graph is the people-side of a context graph: a connected record of the reasoning behind talent decisions—why someone earned a rating, a promotion, or a stretch assignment. That logic usually lives in managers’ heads and calibration debates. An HR context graph makes it durable and searchable.
Context graphs are being built across the enterprise. Regie.ai is building one for sales engagement, PlayerZero for production engineering, Maximor for finance. But the place where reasoning is most invisible—and most consequential—is people decisions.
Glean CEO Arvind Jain captured the challenge: “The why is often a thinking step that usually resides in someone’s head—you can’t actually model it… Over many cycles, those process traces approximate the why.” That’s truer in HR than anywhere. Why did this person get “exceeds expectations” and that one “meets”? Why was she promoted over him? Who’s actually high-potential, and on what evidence? The answers shape careers—and most of them evaporate the moment a calibration meeting ends.
What Your HRIS and Review Tools Miss
HR systems of record store outcomes—final ratings, promotion dates, comp bands. They don’t store the reasoning that produced them: the evidence behind a rating, the precedent that justified a promotion, the cross-tool view of who actually contributed, or the calibration debate that settled a score.
| What HR systems of record capture | What an HR context graph captures |
|---|---|
| The final performance rating | The evidence and reasoning behind it |
| A promotion date and new title | The precedent and trade-offs that justified it |
| The formal org chart | Who actually collaborates, from real work |
| ”Feedback submitted” | What prompted it and how it was weighed |
| Comp band and history | The exception logic and approvals behind each change |
It’s the same gap a sales team has when a discount’s approval lives in a Slack DM—except in HR, the lost context is about people’s careers and a company’s exposure to bias and inconsistency. Without it, every review cycle re-litigates the same questions from scratch, and AI built on top of that record can only repeat the blind spots faster.
How Windmill Builds the HR Context Graph
Windmill builds the HR context graph—a context graph for your people—by capturing the evidence as work happens, and the reasoning behind each people decision as it’s made. Its AI agent, Windy, gathers context year-round across work tools, maps real collaboration, and persists the evidence and rationale behind every rating—turning tribal knowledge into durable, searchable precedent.
Three things make this work. First, Windy gathers context all year across Slack, GitHub, Jira, Asana, and 20+ other tools—so by the time a cycle starts, the evidence behind someone’s performance already exists instead of being reconstructed from memory.
Second, Organizational Network Analysis maps who actually works together, not who reports to whom. Continuous feedback is then requested from real collaborators at the moment a project ships, not months later in a form nobody remembers the context for.
Third, performance reviews and calibrations capture the why, not just the score: the wins behind a rating, the rationale in a calibration session, and the patterns that flag potential bias. The decision and its reasoning land in the same place—instead of one in the HRIS and the other lost in a Slack thread.
Why the HR Context Graph Compounds
An HR context graph compounds because every cycle adds more decision traces. Past reasoning becomes searchable precedent, each new review makes the next one fairer and faster, and the record becomes a defensible answer to “why did we make that call?”—an asset competitors can’t copy.
Models are commoditizing. A high-fidelity record of how your organization actually develops and evaluates people is not. As Aaron Levie put it, “the teams and companies that can accumulate and best utilize context will drive the greatest productivity and highest output.”
None of this requires full autonomy on day one. It starts human-in-the-loop: the agent gathers context, proposes a draft, and records the reasoning while a manager makes the call. Every cycle, the graph grows—and the decisions get faster, fairer, and easier to defend.
The companies that win the next decade of people management won’t be the ones with the best AI models. They’ll be the ones that captured why their best decisions were made. See how this connects to the broader shift toward agent-driven software and year-round performance signals.
Frequently Asked Questions
What is a context graph?
A context graph is a living record of how decisions get made, stitched across systems and time. Coined by Foundation Capital in late 2025, it connects enterprise entities—people, documents, tickets—with the decision traces between them, so AI can see not just what happened, but why. It captures the reasoning that systems of record leave out.
What is an HR context graph?
An HR context graph is the people-side application of a context graph: a connected record of the reasoning behind talent decisions—why someone earned a rating, a promotion, or a stretch project. Today that logic lives in managers' heads, calibration debates, and Slack DMs. An HR context graph makes it durable, searchable, and usable by AI.
Why do AI agents need a context graph?
AI agents act, not just answer. To make good calls without a human in the loop, they need the reasoning behind past decisions, not just the outcomes. Systems of record store the final state; a context graph stores the why. Without it, agents repeat the same mistakes faster—MIT found 95% of enterprise AI pilots deliver no measurable return.
How is a context graph different from an HRIS or system of record?
An HRIS or performance tool stores outcomes: final ratings, promotion dates, comp bands. A context graph stores the reasoning that produced them—the evidence, precedent, trade-offs, and approvals. Systems of record tell you what your org decided; a context graph tells you why, which is what AI needs to act safely.
How does Windmill build an HR context graph?
Windmill's AI agent, Windy, gathers context year-round across Slack, GitHub, Jira, and 20+ other tools, and uses Organizational Network Analysis to map who actually collaborates. Performance reviews and calibrations then capture the reasoning behind each rating and decision—turning talent decisions into durable, searchable precedent instead of tribal knowledge.