What are context graphs? Why do AI agents need them?
Context graphs: how AI agents remember why decisions were made
By<br>Karan
July 01, 2026
Link copied!<br>Copy failed!
Table of contents
Share
Link copied!<br>Copy failed!
In AI agents, a context graph is the part of agent memory that captures decisions.<br>This post explains why AI agents need to capture decisions, how context graphs capture them, how agents use them, and how agents enhance their capabilities and performance with them.<br>An agent failure case<br>It's the last week of the quarter. A renewal agent is working a $480k account. The customer wants 20% off or they walk. The agent's instructions state >$100k accounts should not churn, but the agent's policy caps renewals at 10%. Now what?<br>If a human was handling this, they'll probably use experience and memory to resolve.<br>Didn't we do this exact thing with Globex last quarter? It was a similar story. They were threatening to churn, and someone signed off on 20% because the CEO wanted to retain Fortune 500 logos and the risk was worth taking on a $300k account. It worked and Globex renewed shortly after.
How a person handles it
The rep
How big a discount should the rep approve?
10% cap is clear. Whether Initech justifies an exception is a judgment call.
"Retain all our Fortune 500 accounts. Don't lose one over a renewal discount."
From the CEO, at the last all-hands.
Zoom call on Globex
CC"Accounts this big are worth the risk, and they usually pay."
Thread# renewals
September 12
MC<br>Maya Chen2:41 PM<br>Globex is threatening to walk unless we beat the 10% cap.
3 replies
DR<br>Devang Rao2:44 PM<br>$300k account, Fortune 500, and a great relationship with their VP.
PN<br>Priya Nair2:45 PM<br>Worth retaining. Let's go above the cap for them. I'll handle finance team.
SO<br>Sam Okafor2:47 PM<br>Approved at 20%.
September 29
SO<br>Sam Okafor2:47 PM<br>Globex renewed.
Reply…
Decision: Give 20% discount and close renewal.
Initech renews within 5 days.
This reasoning chain that makes the decision is not written down anywhere your agent can read. The agent will find Globex's exception case in Salesforce, but Salesforce will not tell it that the number was an exception, who approved it, why it was approved, whether the current situation is identical or not.<br>The why lives in -<br>Old slack threads where finance team admits a $300k account is worth the risk.<br>Zoom calls where sales veterans mention these kind of accounts pay eventually.<br>Emails from the CEO saying retaining Fortune-500 logos is critical.<br>These are critical pieces of information needed to make the decision, but the agent cannot access them.<br>So your agent does one of two things -<br>It sends an email informing the policy caps at 10%, and you lose the account.<br>It escalates to a human, who spends 24 hours doing Slack archaeology to reconstruct a decision the company already made once.
How the AI agent handles it
How the AI agent handles it
Your agent
How big a discount can the agent approve?
The 10% cap is clear. Whether Initech justifies an exception is not.
The agent found a different account, Globex, that was given 20% off but does not understand why.
Salesforce, comparable account: Globex
StageRenewal
Discount on record20%
Policy cap10%
Exception rationaleempty
The record shows the number, 20%, but nothing about why it was approved or whether Initech qualifies.
The agent goes looking for the reason.
Slack search
in:#renewals Globex discount
0 results. The thread is older than the 90-day retention window.
Call, AI summary
Pricing and renewal terms discussed
Follow-up scheduled for next week
The reason it was approved was never captured.
Wait for a human
The renewal stalls for about a day
Email: “Policy caps at 10%. We won’t be able to do better.”
Initech churns
Either way, the organization failed to benefit by adopting an AI agent.<br>This is a universal problem. Organizations lose billions each year -<br>making the same mistakes,<br>reinventing the same solutions,<br>wasting time on previously solved problems,<br>being incredibly slow at onboarding new employees,<br>struggling with compliance and audit gaps in an AI-native world.<br>This is because we have gotten extremely good at recording what happened, but we systematically throw away why it happened, which is the one thing your agent needed here.<br>Context graphs store the why in an agent's memory.<br>Foundation Capital called it "a trillion-dollar AI opportunity", which is the kind of phrase that sends people reaching for the back button. But stick around anyway. The term is new and a little overloaded, but it points at something real. If you use or build agents, you'll end up using a context graph soon.<br>Flat context is bad<br>The bad way to give your AI agents context is to just give them all the data, records, documents, rules, policies in a flat context window.<br>Say you have an invoice processing agent. You dump the invoice, PO, vendor record, contract, policy document into the context. Then...