Core Concepts
What are Decision Traces?
Decision traces are the fundamental building blocks of context graphs. They capture not just what decision was made, but the specific context, reasoning, and evidence that led to that outcome.
“Rules tell an agent what should happen in general. Decision traces capture what happened in this specific case.”
— Foundation Capital, “Context Graphs: AI's Trillion-Dollar Opportunity”
Understanding Decision Traces
A decision trace is a structured record of a decision event that includes:
- ●The decision itself — what outcome was chosen
- ●The context — what information was available at the time
- ●The reasoning — why this option was selected over alternatives
- ●The actors — who (or what) made or approved the decision
- ●The timestamp — when the decision was made
- ●The evidence — supporting data, policies, or precedents referenced
Decision Trace Example
Consider a pricing exception request in an enterprise sales process:
{
"trace_id": "dt_2025_01_07_pricing_001",
"decision_type": "pricing_exception",
"outcome": "approved",
"timestamp": "2025-01-07T14:32:00Z",
"context": {
"customer": "Acme Corp",
"customer_tier": "enterprise_tier_1",
"deal_size": "$2.4M ARR",
"requested_discount": "15%",
"standard_discount_max": "10%"
},
"reasoning": [
"Customer is Tier 1 with 3-year history",
"Q4 revenue target pressure",
"Competitive displacement opportunity",
"Similar exception approved for Beta Inc (dt_2024_09_pricing_042)"
],
"actors": {
"requester": "sales_rep_jane",
"approver": "vp_sales_michael",
"approval_level": "VP override"
},
"evidence": {
"policies": ["pricing_policy_v3.2"],
"precedents": ["dt_2024_09_pricing_042", "dt_2024_06_pricing_018"],
"attachments": ["competitive_analysis.pdf"]
}
}This decision trace captures not just that a discount was approved, but why it was approved, who approved it, and what precedents were referenced. When an AI agent encounters a similar situation, it can search for this trace and understand how to handle the case.
Why Decision Traces Matter
For AI Agents
Decision traces provide the institutional memory AI agents need to handle edge cases. Instead of failing on every exception, agents can search for similar past decisions and apply the same reasoning.
For Compliance
Decision traces create a complete audit trail of how and why decisions were made. This is essential for regulatory compliance (EU AI Act, SOX, GDPR) and internal governance.
For Knowledge Preservation
When experienced employees leave, their decision-making knowledge often leaves with them. Decision traces preserve this tribal knowledge in a searchable format.
For Continuous Improvement
Decision traces enable analysis of past decisions to identify patterns, improve policies, and train better AI models. Every decision becomes a learning opportunity.
Rules vs Decision Traces
Traditional systems rely on explicit rules. Decision traces capture what happens when rules aren't enough:
| Rules | Decision Traces |
|---|---|
| "Max discount is 10%" | "15% approved because of Tier 1 + competitive threat" |
| "Refunds within 30 days" | "45-day refund approved due to shipping delay we caused" |
| "Requires manager approval" | "VP override because manager was on leave" |
Rules define the default behavior. Decision traces capture the exceptions—and in enterprise operations, exceptions are where the real complexity lives.
Frequently Asked Questions
What is a decision trace?
A decision trace is a record that captures not just what decision was made, but the specific context, reasoning, exceptions, and approvals that led to that outcome. It documents the "why" behind each decision.
Why are decision traces important for AI agents?
Decision traces provide AI agents with the institutional memory and precedent they need to handle edge cases and exceptions. Without them, AI agents can only follow explicit rules and fail on unwritten exceptions.
How are decision traces different from logs?
Logs record what happened (events). Decision traces record why it happened (reasoning). A log might say "discount applied: 15%". A decision trace explains why 15% was chosen over the standard 10%.