KPI 05 of 07 · Relationship Phase · Consumer-side · Lower is better
Trust Erosion Index
The rate at which human trust in an agent system is declining over time
Abbreviation: TEI · Inverse metric (lower is better)TEI is the relationship health metric of Agentic Experience Design. Trust is the primary material of AXD - the medium in which all human-agent relationships are built. TEI measures whether that material is holding or crumbling. A rising TEI is the most dangerous signal in an agentic system, because trust erosion is a leading indicator of delegation abandonment.
Trust erosion is not a single event. It is the cumulative effect of many small disappointments, unexpected outcomes, opaque decisions, and unacknowledged failures. Each trust-damaging event adds to the erosion. Each trust-building event (successful delegation, transparent reasoning, proactive disclosure) reduces it. TEI captures the net direction of this ongoing negotiation.
The metric is deliberately framed as erosion rather than trust level. Trust level is difficult to measure directly - it is subjective, contextual, and varies across individuals. Erosion is observable through behavioural signals: users adding more constraints, checking outcomes more frequently, reducing delegation scope, or abandoning the agent entirely. These behavioural changes are the measurable surface of trust dynamics.
TEI draws from the Trust Architecture pillar of the Four Pillars of AXD Readiness and the Trust Calibration Framework. It operationalises the founding principle that trust is the primary material of AXD - and that the designer's responsibility is to maintain, repair, and strengthen that material over time.
Trust architecture and erosion dynamics
TEI is governed by the trust architecture of the agentic system. Systems with well-designed trust recovery mechanisms will have lower TEI even when trust-damaging events occur, because the recovery pathways prevent individual failures from compounding into systemic erosion.
The asymmetry of trust dynamics is critical: trust is built slowly through consistent positive experiences but can be destroyed rapidly by a single severe violation. TEI captures this asymmetry. A single unexpected high-value purchase by an agent can produce more trust erosion than ten successful low-value purchases can build. Designing for this asymmetry is the core challenge of trust architecture.
Numerator
Weighted trust-damaging events minus weighted trust-building events
Denominator
Total trust-relevant interactions in the measurement period
× 100 = TEI %
Inverse metric: lower is better. A negative TEI indicates net trust growth. Weight trust-damaging events more heavily than trust-building events to reflect the asymmetry of trust dynamics.
Measurement protocol
Identify trust-relevant events in the human-agent interaction: delegation failures, constraint violations, unexpected outcomes, opaque decisions (trust-damaging) and successful delegations, transparent reasoning, proactive disclosure, effective recovery (trust-building). Weight each event type based on severity and impact.
Supplement event-based measurement with behavioural signals: changes in delegation scope (narrowing indicates erosion), changes in constraint density (increasing indicates erosion), changes in outcome checking frequency (increasing indicates erosion), and delegation abandonment (the terminal signal).
Report TEI as a rolling average over a 30-day window. Short-term spikes are normal after individual trust events; the trend is what matters. A sustained upward trend in TEI requires immediate attention to trust architecture design.
Four levels of trust health (lower is better)
>40%
Rapid trust collapse. Users are actively reducing delegation scope or abandoning the agent. Multiple trust violations have occurred without adequate recovery. Immediate intervention required in trust architecture design.
20-40%
Noticeable trust decline. Users are becoming more cautious with delegations, adding more constraints, or checking outcomes more frequently. Trust recovery mechanisms are insufficient.
5-20%
Minor trust fluctuations within normal operating range. Some trust events are occurring but recovery mechanisms are functioning. Users maintain delegation scope with minor adjustments.
<5%
Trust is stable or growing. Trust violations are rare and well-recovered. Users are expanding delegation scope over time. Trust architecture is functioning as designed.
What moves TEI up, down, and sideways
Transparent agent reasoning, proactive uncertainty disclosure, designed trust recovery pathways, consistent outcome quality, appropriate human re-engagement triggers, clear constraint acknowledgment.
TEI is an inverse metric - lower is better. A rising TEI is a leading indicator of delegation abandonment. By the time users stop delegating, the trust relationship is already severely damaged. Monitor TEI trends weekly.
Opaque agent decision-making, unexpected outcomes without explanation, constraint violations without acknowledgment, absence of trust recovery mechanisms, over-promising and under-delivering on agent capabilities.
Why TEI matters commercially
TEI is the leading indicator of agentic commerce sustainability. High TEI means consumers are losing trust in agent-mediated purchasing. When trust erodes sufficiently, consumers revert to manual purchasing - eliminating the entire agentic commerce channel. TEI predicts whether the agentic commerce channel will grow or collapse.
For the industry as a whole, aggregate TEI across all agent platforms determines the pace of agentic commerce adoption. If consumers collectively experience high trust erosion, the entire market slows. Individual businesses and agent platforms that maintain low TEI will capture disproportionate market share as trust-eroded consumers migrate to more reliable alternatives.