AI handles realistic customer calls, follows policy, and uses database tools.
Available on Request · Sample Available
Voice-Agent Tasks (𝜏²)
Realistic customer calls, policy adherence, and database-tool use.
Volume
0
Volume
Scalable Volume
Coverage
Agentic Workflows
0.0
%
Gemini 3.5 Flash
Targets
Tau2
Modality & Delivery
Text→Text
Formats
JSON
What You Receive
Coverage
Package Structure
- data/
- db.json
- policy.md
- split_tasks.json
- tasks.json
- transcripts/
- task-1.txt
- task-2.txt
- task-3.txt
- task-4.txt
- task-5.txt
Differentiators
- Runnable insurance tasks combine policy following, database tools, and final-state grading.
Record Schema
Results
| Model | Metric | Result |
|---|---|---|
| Gemini 3.5 Flash | Pass rate |
Methodology & Quality
Methodology
SME involvement
Domain-expert curated with up to L2 review
Stressors
Customer interaction, policy compliance, tool use, and database-state accuracy.
Design criteria
- Score task completion by comparing the final database with the verified correct result.
Build assumptions
Quality approach
- Automated final-database comparison against a verified gold state.
Known limitations
Customization & Scalability
Custom sets available. Additional volume in progress