Multifamily Risk Adjustment Workflow with Mastra AI and GPT-5 Pro
Safety investments made by commercial property owners often fail to translate into premium reductions due to fragmented underwriting workflows. You will orchestrate an automated risk assessment pipeline using Mastra AI and GPT-5 Pro to evaluate IoT sensor installations and recalculate property premiums. The system must produce audited premium adjustments within a 5% margin of target actuarial tables.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an IoT safety investment and underwriting adjustment workflow using Mastra AI and GPT-5 Pro reasoning models.
How work is evaluated
Evaluates accuracy of calculated premium discounts based on property IoT safety features.
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.
within_margin
Validates discount percentage is within 1% absolute margin of target.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
actuarial_mae
Mean Absolute Error of calculated premium vs ground truth • target: 250 • range: 0-5000
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Build Mastra AI workflow graphs connecting sensor telemetry parsers to underwriting engines.
Incorporate GPT-5 Pro high-reasoning prompts to evaluate IoT risk mitigation credit validity.
Validate pricing model outputs against historical commercial property loss databases.
Reference links and supporting material
Dataset of 150 commercial multifamily property profiles containing IoT hardware logs, maintenance history, and manual underwriting decisions.
Challenge setup
Evaluates accuracy of calculated premium discounts based on property IoT safety features.
Challenge input
JSON object with property_id, iot_devices (array), base_premium, historical_losses
Mastra AI
Assigned framework for workflow graph implementation.
GPT-5 Pro
Assigned high-reasoning model for underwriting evaluation.
Evaluated output
JSON object with adjusted_premium, discount_percentage, credit_breakdown, rationale
- Validates discount percentage is within 1% absolute margin of target.
- Mean Absolute Error of calculated premium vs ground truth • target: 250 • range: 0-5000
- Benchmark: Underwriting Workflow Precision Benchmark
- Actuarial Mae target: 250
- 1 public reference case
- JavaScript execution harness
- javascript 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
- GPT-5 Pro · Required
- Mastra AI · Optional
- GPT-5 · Optional
- Mastra AI · Optional
Evaluation contract
- within_margin · Weight 1
- actuarial_mae · Weight 1
Metrics
- actuarial_mae · Target 250
Benchmarks
- Underwriting Workflow Precision Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON object with property_id, iot_devices (array), base_premium, historical_losses
Output: JSON object with adjusted_premium, discount_percentage, credit_breakdown, rationale
Execution harness: JavaScript
Evaluation tools
- Mastra AI · Agent & Workflow Framework
- GPT-5 Pro · LLM Reasoning Engine
Recipe freeze marker
No recipe freeze time is recorded.