Active Started Nov 1, 2024
Voice Agent v2.1
Second-generation voice AI with improved latency and natural conversation flow. Focuses on reducing response time while maintaining quality.
Progress 75%
Innovation Lab
What We're Building
AI, automation, and customer experience experiment records. Status labels reflect the recorded stages; the latest update in this archive is dated February 3, 2025.
Latest recorded update: Feb 3, 2025
3
Active
2
Shipped
3
Learned
Latest Update
Feb 3, 2025
Voice latency optimization hits target
Reduced p95 latency from 450ms to 280ms by implementing edge caching for model inference. Now testing with 5 production clients.
Update Archive
Update Archive
-
Feb 3, 2025
Shipped Voice latency optimization hits target
Reduced p95 latency from 450ms to 280ms by implementing edge caching for model inference. Now testing with 5 production clients.
-
Jan 27, 2025
Testing Multi-speaker diarization enters testing
Our speaker identification model can now distinguish 4+ speakers in conference calls with 94% accuracy. Preparing for client pilots.
-
Jan 20, 2025
Failed Real-time translation v1 paused
Latency too high for voice calls (800ms+ roundtrip). Pivoting to async translation for email/chat while we explore edge deployment.
Active Experiments
Active Experiments
Testing Started Dec 1, 2024
Multi-Speaker Diarization
Automatic speaker identification and separation in conference calls. Essential for accurate call summaries and action items.
Progress 85%
Testing Started Oct 15, 2024
Real-Time Sentiment Analysis
Detect customer frustration, confusion, or satisfaction in real-time to guide agent responses and escalation.
Progress 60%
Shipped
Shipped
Experiments that graduated to production
Intent-Based Smart Routing
Automatically route inquiries to the best-matched agent based on detected intent, sentiment, and agent expertise.
Shipped · Oct 15, 2024
Automated QA Scoring
AI-powered quality assurance that scores 100% of interactions instead of random sampling.
Shipped · Aug 1, 2024
The Graveyard
The Graveyard
Failed experiments and what we learned. We share these openly because failure is part of innovation.
Failed · Apr 15, 2024
Fully Autonomous Chatbot
Edge cases and emotional situations require human judgment. Pivoted to human-AI hybrid approach.
Failed · May 1, 2024
Agent Voice Cloning
Legal and ethical concerns around consent. Customers preferred knowing when they talked to AI vs humans.
Failed · Mar 15, 2024
Predictive Volume Modeling v1
Too many external variables (marketing campaigns, product launches) made predictions unreliable. Now using shorter 7-day windows.
start
Want to Be a Beta Tester?
We partner with forward-thinking companies to test new capabilities before general release.