Summary

Enhance AIR Analytics to provide comprehensive visibility into Unresolved Calls, including categorization, metrics, and actionable insights across all failure types: Abandoned Calls, Dropped Calls, Lost Calls, Out-of-Scope Intent, Multilingual/Dialect Failure, and other system-level or user-level breakdowns.

Problem / Opportunity

Currently, AIR Analytics offers limited insight into why certain calls go unresolved. Without detailed categorization and metrics, users cannot accurately diagnose performance issues, optimize call flows, or understand where the AI receptionist is failing. This diminishes the ability to improve customer experience and reduces confidence in the system’s accuracy and coverage.

Proposed Solution

Introduce a dedicated Unresolved Calls Analytics Module that:

1. Categorizes Unresolved Calls

Break down unresolved calls into clearly defined types such as:

Abandoned Calls – caller hangs up before AIR begins or completes an action

Dropped Calls – technical failure causing an unintended disconnect

Lost Calls – system timeout or call routing failure

Out-of-Scope Intent – caller intent not recognized or unsupported

Multilingual / Dialect Failure – failed understanding due to language, accent, or dialect

Other Failures – e.g., transcription error, ASR failure, internal error

Each call should be assigned a primary failure reason with optional secondary tags.

1. Provide Detailed Metrics for Each Category

Include:

Total number of unresolved calls

Percent of total call volume

Trend over time (daily, weekly, monthly)

Comparison against resolved calls

Duration before failure

Repeat callers (if applicable)

1. Surface Actionable Insights

Examples:

High rates of Out-of-Scope Intent → recommendation to expand training data or intents

Spike in Dropped Calls → possible telephony provider issue

Pattern in multilingual failures → need for improved language model coverage

Abandoned call peaks → user experience or menu structure review

1. Include Call Playback + Transcription (if available)

Allow users to review unresolved call recordings (configurable for compliance) to understand root causes.

1. Export & API Access

Exportable CSV and API endpoints to ingest unresolved call data into BI tools.

Business Impact

Improves accuracy and performance of the AI receptionist

Reduces missed business opportunities due to unresolved calls

Enables faster troubleshooting and reduces support burden

Helps product teams prioritize model improvements and training needs

Enhances customer trust by offering transparency and actionable insights

Success Metrics

Reduction in unresolved call rate by X% over Y months

Increase in first-intent recognition rate

Decrease in average time to diagnose call flow issues

Improved customer satisfaction score (CSAT) around call handling

Additional Notes

Optional enhancements:

Heatmaps for peak unresolved call times

Comparison by location or phone number

Ability to assign custom tags to unresolved calls

Alerts for anomaly detection (e.g., sudden spike in dropped calls)