This repository is a working collection of experiments exploring practical applications of artificial intelligence and AI agents.
The goal is to learn by building: identifying real workflows, testing where agents can add value, and documenting both their capabilities and their limitations.
| Project | Description | Status |
|---|---|---|
| Agentic Reserving | An experiment using AI agents to support the actuarial loss-reserving process. | In progress |
Additional experiments will be added as the repository develops.
These projects may examine questions such as:
- Which professional workflows can agents complete reliably?
- Where do agents still require expert judgment or review?
- Can agentic workflows make technical analysis more reproducible and transparent?
- How should agents validate their inputs, calculations, and conclusions?
- What controls are needed before these tools can be used in consequential decisions?
Each experiment lives in its own folder and has a dedicated README describing:
- the problem being explored;
- the proposed agentic workflow;
- setup and usage instructions;
- outputs and examples;
- known limitations; and
- ideas for future development.
This is an experimental repository. The projects may change substantially as I test different models, tools, prompts, and agent architectures.
The materials in this repository are for experimentation and education. They should not be treated as production systems or as substitutes for professional judgment. Outputs should be independently reviewed before being used for business, financial, actuarial, or regulatory decisions.