How to Measure AI Agent Impact in BoldDesk
BoldDesk AI Agents reduce live‑agent chat volume by resolving customer questions through automated self‑service. You can measure this effectiveness using the Deflection Rate in the AI Agent Performance Dashboard, which shows how often AI Agents handle (or customers abandon) conversations without escalation to a human agent.
This guide explains how to track AI self‑service effectiveness using the chat conversation deflection rate metric.
What “Chat Conversation Deflection Rate” measures
The Deflection Rate indicates how many conversations were completed without human‑agent involvement specifically, conversations handled by the AI with no escalation to a human.
Operational signals this metric helps you monitor
A stronger deflection rate typically suggests:
- More conversations handled by AI without escalation (AI self‑service is working).
- Fewer transfers to human agents (lower escalation pressure).
- Reduced manual workload (tracked as “Human Hours Saved” in the dashboard).
How the metric is calculated (practical interpretation)
Because BoldDesk defines Deflection Rate as conversations handled or abandoned by the AI without human escalation, you can interpret it as:
- Deflection Rate (%) = [(AI‑handled conversations + AI‑abandoned conversations) ÷ Total conversations] × 100
Access prerequisites
To view the AI Agent Performance Dashboard, the user must be able to access the Reports module in the Agent Portal. Reports and dashboards are available to agents with the appropriate permissions.
If a user can’t access the dashboard, verify their reporting permissions (Reports access) in BoldDesk administration.
How to track AI Agent chat deflection
1) Open the AI Agent Performance Dashboard
-
In the Agent Portal, go to Reports.
-
Open AI Agent Performance Dashboard.
-
Use the Performance Overview tab for trend tracking and headline metrics.
The dashboard includes key AI outcomes like Total Conversations, Deflection Rate, Escalation Rate and AI-Handled Hours.
2) Review “Deflection Rate” and supporting metrics
In the Overview Statistics widget, review:
- Total Conversations (all conversations created/initiated in the selected period).
- Deflection Rate (conversations handled or abandoned by AI with no escalation).
- Escalation Rate (conversations transferred from AI to a human agent).
- Human Hours Saved (estimated time saved due to AI handling).
3) Use “Conversation Breakdown” to find what drives deflection (or escalations)
Use the dashboard’s Conversation Breakdown analysis to segment conversations by attributes such as:
-
Handling type
-
Brand
-
Handling category
-
Source / channel
This helps you identify which channels/topics are deflecting well and which are escalating more often.
How to interpret “Deflection Rate” results
Healthy pattern: High Deflection Rate with manageable Escalation Rate
This typically indicates:
- AI Agents are successfully handling conversations without needing a human handoff.
- Customers can complete common chat interactions without requesting a live agent.
Concerning pattern: Deflection drops and Escalation rises
This typically indicates:
- The AI Agent is encountering topics it can’t resolve confidently (coverage gaps).
- Customers are choosing Talk to a Person, which escalates the conversation to a human agent.
- Agents are taking over conversations more often (agent takeover contributes to escalation).
What to watch when improving deflection
- Identify which AI Agents have the highest escalations using Top 5 Escalated Agents (useful for prioritizing refinement).
Frequently Asked Questions (FAQs)
-
Can I identify what causes escalations (for example, channel or topic)?
Yes. The dashboard supports conversation breakdown analysis by attributes such as handling type, brand, handling category, and source/channel. -
What counts as an escalated conversation?
Escalated conversations are those transferred from the AI Agent to a human agent for further handling. This can occur when a customer requests a human (for example, using Talk to a Person) or when the conversation is taken over by an agent. -
Is there a “good” deflection rate target?
There’s no single universal target. The most reliable approach is to track trend improvement over time for the same brand/channel/timeframe and compare against your baseline before AI Agent deployment. The dashboard is designed for trend tracking and performance optimization.
