KPI 04 of 07 · Execution Phase · Consumer-side · Higher is better
Delegation Completion Rate
The percentage of delegated tasks completed to human satisfaction without intervention
Abbreviation: DCRDCR is the execution metric of Agentic Experience Design. It measures the most fundamental question of the human-agent relationship: when you delegate a task to an agent, does it get done? Not just completed in a technical sense, but completed to your satisfaction, within the constraints you specified, producing the outcome you intended.
This is the first consumer-side metric in the AXD Metrics Standard. While AIR, AACR, and CSAS measure the merchant's readiness for agentic commerce, DCR measures the quality of the delegation relationship between a human and their agent. It sits at the intersection of intent specification, constraint design, and agent capability.
DCR is not simply a task success rate. A task that completes but violates a constraint is a DCR failure. A task that completes but produces an outcome the human did not intend is a DCR failure. A task that requires human intervention to complete - even if the final outcome is satisfactory - is a DCR failure. The metric is deliberately strict because delegation trust depends on reliability, not on occasional success.
The metric draws directly from the Delegation Design pillar of the Four Pillars of AXD Readiness and the Intent Architecture Framework. Poor DCR is almost always a design problem, not a capability problem. The agent fails not because it cannot execute, but because the delegation was poorly specified, the constraints were unclear, or the recovery pathways were absent.
How delegation design governs DCR
DCR is governed by the quality of the delegation contract between human and agent. This contract includes the intent specification (what the human wants), the constraint boundaries (what the agent must not do), the success criteria (how completion is defined), and the recovery pathways (what happens when things go wrong).
Commerce protocols influence DCR indirectly. When merchants provide clear, structured product data through ACP or UCP, agents can make better purchasing decisions on behalf of their humans. When inventory is real-time and pricing is consistent, agents encounter fewer failure modes during execution. The merchant's protocol readiness directly affects the consumer's delegation success.
Numerator
Delegated tasks completed to human satisfaction without intervention
Denominator
Total tasks delegated to the agent
× 100 = DCR %
Report alongside delegation scope complexity to prevent artificial inflation through narrow task selection.
Measurement protocol
Track every delegation event - the moment a human grants authority to an agent to act. Record the intent specification, constraints, and success criteria at the point of delegation. Track the execution through to completion or failure. Classify the outcome as: completed to satisfaction, completed with constraint violation, completed with human intervention, or failed.
Only "completed to satisfaction without intervention" counts as a DCR success. All other outcomes are DCR failures, even if the final result was acceptable. This strict definition ensures that DCR measures true delegation reliability rather than eventual task completion.
Segment DCR by delegation complexity (simple, moderate, complex), by domain (purchasing, scheduling, research), and by agent platform. This segmentation reveals where delegation design is mature and where it needs improvement.
Four levels of delegation reliability
<30%
Most delegated tasks fail or require human takeover. Intent specification is likely too vague, constraint boundaries are unclear, or the agent lacks the capability to execute the delegated scope. Fundamental delegation design issues.
30-60%
Moderate completion. Simple delegations succeed but complex or multi-step tasks fail. The agent can handle well-defined tasks but struggles with ambiguity, edge cases, or scope that requires judgment.
60-85%
High completion across standard delegation types. Failures are concentrated in edge cases and novel situations. The delegation design is mature for common scenarios but needs refinement for exceptions.
>85%
Near-complete delegation success. The agent handles standard and edge cases reliably. Failures are rare and well-handled through designed recovery pathways. Trust calibration is mature.
What moves DCR up, down, and sideways
Clear intent specification with explicit constraints, well-defined success criteria, designed recovery pathways for common failure modes, progressive delegation that builds capability over time, outcome-based rather than process-based delegation.
DCR can be artificially inflated by limiting delegation scope to only simple tasks. A high DCR with narrow scope is less valuable than a moderate DCR with broad scope. Always report DCR alongside delegation scope complexity.
Vague intent specification, missing constraint boundaries, no recovery pathways for failure modes, over-delegation beyond agent capability, absence of human re-engagement triggers for edge cases.
Why DCR matters commercially
DCR determines whether consumers will continue to delegate to agents. A consumer whose agent consistently fails to complete delegated tasks will stop delegating - reverting to manual purchasing, manual scheduling, manual research. Low DCR kills agentic commerce adoption at the consumer level.
For merchants, DCR matters indirectly but powerfully. Merchants whose commerce infrastructure supports high DCR - through clear product data, reliable inventory, and protocol integration - will be preferred by agents that learn from delegation outcomes. High-DCR merchants become default recommendations because agents learn that delegations involving those merchants succeed.