Build Voice-Enabled Lead Qualification Agent using OpenAI Agents SDK & Resemble AI
As homebuilders face mounting interest rate pressures, converting dynamic buyer inquiries into verified sales is critical. This challenge requires building an autonomous voice-enabled lead qualification agent using the OpenAI Agents SDK and Resemble AI voice synthesis. The agent interacts with interested home buyers over voice channels, dynamic conversational state, and interest-rate sensitivity calculators. It negotiates rate buy-down incentives, calculates monthly mortgage payments under prevailing mortgage rates, and schedules on-site visits with sales teams.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop a conversational voice agent with OpenAI Agents SDK and Resemble AI to qualify buyer interest and offer rate buy-down calculations.
How work is evaluated
Evaluates lead qualification accuracy, payment calculation correctness, and synthetic voice prompt response generation.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
Buydown Calculation Accuracy
Validates year 1 buydown payment calculation within $1 accuracy
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Qualification Accuracy
Correct determination of buyer qualification based on DTI ratio (<43%) • target: 1 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement interactive multi-turn dialogue management using OpenAI Agents SDK
Integrate Resemble AI for ultra-realistic voice output during buyer qualification calls
Build rate-buydown tool logic to calculate 2-1 buy-down affordability during the conversation
Measure buyer intent score and trigger high-priority CRM handoffs
Reference links and supporting material
Dataset of 100 buyer inquiries including annual income, home target budget, and credit score tiers.
Challenge setup
Evaluates lead qualification accuracy, payment calculation correctness, and synthetic voice prompt response generation.
Challenge input
JSON with home_price, down_payment_pct, rate, buyer_income
OpenAI Agents SDK
Required framework for task orchestration.
Resemble AI
Required tool for streaming realistic AI voice.
Evaluated output
JSON with qualified (boolean), initial_monthly_payment, buydown_year1_payment, response_text
- Validates year 1 buydown payment calculation within $1 accuracy
- Correct determination of buyer qualification based on DTI ratio (<43%) • target: 1 • range: 0-1
- Benchmark: Voice Conversion Bench
- Qualification Accuracy target: 1
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Resemble AI · Required
- OpenAI · Optional
- Patronus AI · Optional
- Patronus AI · Optional
Evaluation contract
- Buydown Calculation Accuracy · Weight 1
- Qualification Accuracy · Weight 1
Metrics
- Qualification Accuracy · Target 1
Benchmarks
- Voice Conversion Bench
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON with home_price, down_payment_pct, rate, buyer_income
Output: JSON with qualified (boolean), initial_monthly_payment, buydown_year1_payment, response_text
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · Agent Framework
- Resemble AI · Voice Synthesis
Recipe freeze marker
No recipe freeze time is recorded.