Daniella's Customer Brain

Honeycomb <> OneAI: Weekly Sync

3 min read

Honeycomb <> OneAI: Weekly Sync

Date: 2026-01-20 | Duration: 36.02000045776367 min | Account: honeycomb

External attendees: jan.j@honeycombinsurance.com, jon@honeycombinsurance.com, rafael@honeycombinsurance.com

Summary

The team discussed improvements to the four-way triage system, achieving 70% accuracy in flagging issues, with Amit Ben acknowledging the need for manual reviews to address false positives caused by system limitations. Rafael Mintz raised concerns about leads being disqualified without clear rejection, advocating for manual requalification of ambiguous cases to retain potential prospects. The team identified technical bugs that led to inconsistent call handling and agreed to refine the insurance product knowledge base for better AI understanding. Voice agent performance was also reviewed, resulting in Raquel being set as the default voice for calls after Becca underperformed. Continuous monitoring of triage accuracy and real-time alerts were emphasized to ensure ongoing improvements in call quality and user qualification.

Key Points

🔍 Triage System Updates: 70% of triage flags now correctly identify issues; false positives reduced to 4, still requiring manual review. 🔄 Lead Management Concerns: Ambiguous cases are flagged for manual requalification to avoid losing valid leads during triage. 📚 Knowledge Base Improvements: Explicit listings for covered insurance types are added to enhance AI understanding and qualification flow. 🎤 Voice Agent Changes: Raquel replaced Becca as the default voice, improving engagement based on data from over 785 calls. 📞 Call Sequencing Adjustments: Manual triage focus on same-day calls ensures prompt issue detection and better funnel management. 📈 Monitoring Progress: Continuous tracking of false positives is essential for maintaining call quality and ensuring responsiveness to issues.

Overview

  • Triage System Updates: 70% of triage flags now correctly identify issues; false positives reduced to 4, still requiring manual review.

  • Lead Management Concerns: Ambiguous cases are flagged for manual requalification to avoid losing valid leads during triage.

  • Knowledge Base Improvements: Explicit listings for covered insurance types are added to enhance AI understanding and qualification flow.

  • Voice Agent Changes: Raquel replaced Becca as the default voice, improving engagement based on data from over 785 calls.

  • Call Sequencing Adjustments: Manual triage focus on same-day calls ensures prompt issue detection and better funnel management.

  • Monitoring Progress: Continuous tracking of false positives is essential for maintaining call quality and ensuring responsiveness to issues.

Action Items

Amit Ben Review and update triage prompt to reduce false positives and ensure clear triage logic (05:05) Document and explicitly include “habitational” and “commercial apartment” insurance in AI knowledge base and FAQ for Tyler (17:14) Investigate causes of repeated calls after disqualification and why qualification flow was interrupted in specific call cases (30:00) Disable “Becca” voice persona and ensure “Raquel” is the default Tyler voice (26:54) Monitor and track triage flags from current day onwards, prioritize new triage alerts and close out stale issues (34:24) Note the need to make Tyler AI’s responses more human-like by shortening the reading of full names at caller greetings (27:08)

Rafael Mintz Continue sending daily triage reports to Amit and perform detailed triage reviews (34:20) Provide feedback on AI behavior and triage interaction nuances, especially around ambiguous disqualification leads (28:14) Evaluate transitions in conversation flow and qualification process in Tyler AI to ensure smooth lead conversion paths (19:48)

Daniella Block Assist in integrating step count tracking and reintroduce daily meeting treadmill walking exercise for wellbeing focus (03:00)

Graph View