Daniella's Customer Brain

Honeycomb <> OneAI: Weekly Sync

3 min read

Honeycomb <> OneAI: Weekly Sync

Date: 2026-02-03 | Duration: 26.729999542236328 min | Account: honeycomb

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

Summary

The meeting centered on enhancing the accuracy of the AI model and call classification, leading to a significant reduction in misclassifications. Amit Ben reported successful retraining of the AI model, correcting 10 previous errors and improving call flow by distinguishing user intent more effectively. The team identified that 44% of classification issues stemmed from users stating they had already received a quote, prompting adjustments to script language to avoid premature call terminations. They also discussed refining how call statuses are communicated, implementing flags for previous statuses to prevent false positives. Additionally, the team aligned on qualification logic and exceptions to ensure accurate customer experience management. The collaboration aims to maintain high-quality leads and data accuracy, with ongoing reviews and transparent communication.

Key Points

✅ AI Model Update: Retaining 10 AI misclassifications resolved; accuracy improved significantly after new training. 📞 Call Classification Issues: 44% of issues stemmed from ‘already got a quote’; adjusted script increased user intent clarity. 🔍 Call Status Communication: Implementing flags for ‘previous call status’ will reduce false positives and enhance data reliability. 📝 Qualification Logic Refinement: Only 5 errors out of 40 calls indicate strong audit accuracy; new questions address specific scenarios. 🔄 Process Review Commitment: Sharing detailed call audits will streamline issue identification and improve ongoing qualification model.

Overview

  • AI Model Update: Retaining 10 AI misclassifications resolved; accuracy improved significantly after new training.
  • Call Classification Issues: 44% of issues stemmed from “already got a quote”; adjusted script increased user intent clarity.
  • Call Status Communication: Implementing flags for “previous call status” will reduce false positives and enhance data reliability.
  • Qualification Logic Refinement: Only 5 errors out of 40 calls indicate strong audit accuracy; new questions address specific scenarios.
  • Process Review Commitment: Sharing detailed call audits will streamline issue identification and improve ongoing qualification model.
  • Meeting Coordination: Rafael traveling next week; John may lead, with open communication for urgent updates during absences.

Action Items

Amit Ben Provide Honeycomb a flag for previous call status to enable ignoring or special handling of repeat or non-regression updates (10:28) Investigate and implement technical solution to send previous call results via webhook or API for newer contacts (10:54) Share spreadsheet detailing audited call analysis and flagged false positives for continued review (13:37) Implement HOA insurance follow-up question to differentiate eligibility per Rafael’s phrasing (19:23) Add optional flag to mark contacts qualified through the not-competitive price fallback script for Honeycomb’s logic (23:56)

Rafael Mintz Develop internal logic to ignore status updates where contact is already in a qualified, disqualified, or converted state to reduce funnel errors (11:47) Continue forward flagging and sharing of suspicious or questionable call transcripts for Honeycomb AI audit review (14:24) Update qualification logic to handle “not competitive” but still qualified scenarios as per new AI script and flag from Amit (23:31) Coordinate with John to potentially lead next weekly sync or confirm cancellation based on agenda and travel plans (24:57)

Daniella Block Support communication and scheduling coordination during upcoming meetings and follow-ups (00:02)

Graph View