Skip to main content
Back to challenges
Challenge

OpenAI Agents SDK Agent Copilot for Dynamic Policy Structuring

Insurance brokers need AI copilots that assist human agents in structuring complex multi-peril coverage options in real time. Build an OpenAI Agents SDK copilot with dynamic tool execution to recommend coverage add-ons during policy quoting calls.

Machine LearningHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Implement an OpenAI agent featuring context preservation and function calling to generate personalized insurance proposals.

Delivery guide

How work is evaluated

Evaluation

Evaluates coverage recommendation relevance and latency during simulated customer intake calls.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

umbrella_recommendation_check

Ensures Personal Umbrella is recommended when pool is mentioned

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

endorsement_recall

Recall metric of recommended coverage options • target: 0.95 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Develop an OpenAI Agents SDK copilot with live tool registration for coverage recommendations

  • Maintain multi-turn transcript context during sales calls to extract client risk parameters

  • Generate dynamic coverage comparison quotes in under 2 seconds

Resources and assets

Reference links and supporting material

Dataset notes

Transcripts of 150 intake conversations between insurance agents and policyholders.

Challenge setup

Evaluates coverage recommendation relevance and latency during simulated customer intake calls.

Challenge input

JSON containing customer_call_transcript

OpenAI Agents SDK

Official framework for building conversational agent tools and copilots

Evaluated output

JSON with recommended_endorsements, estimated_additional_premium, explanation

Evaluation checks
  • Ensures Personal Umbrella is recommended when pool is mentioned
  • Recall metric of recommended coverage options • target: 0.95 • range: 0-1
Configured evaluation evidence
  • Benchmark: Copilot-Rec-Bench
  • Endorsement Recall target: 0.95
  • 1 public reference case
Configured runtime
  • 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

Action Space
  • OpenAI · Required
  • Cohere Platform · Optional
Safety / Guardrails
  • Protect AI · Optional

Evaluation contract

  • umbrella_recommendation_check · Weight 1
  • endorsement_recall · Weight 1

Metrics

  • endorsement_recall · Target 0.95

Benchmarks

  • Copilot-Rec-Bench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing customer_call_transcript

    Output: JSON with recommended_endorsements, estimated_additional_premium, explanation

    Execution harness: Python

Evaluation tools

  • OpenAI Agents SDK · Agent SDK

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about OpenAI Agents SDK Agent Copilot for Dynamic Policy Structuring