Project Scope
Agentic Engineering & Data Modernization
Prepared for: Evergreen
Prepared by: Evan Fischell Consulting
Date: September 4, 2026
Work with Evergreen's team to establish governed agentic engineering and implement a modern data and reporting platform.
Three months. Three phases.
| Phase | Timing |
|---|---|
| 1 — Agentic Engineering & CI/CD | Month 1 |
| 2 — Data Platform & Ingestion | Month 2 |
| 3 — Agentic-First Reporting | Month 3 |
Phase 1 — Agentic Engineering & CI/CD
Duration: one month
What we will do
- Evaluate token broker options (Azure Foundry, OpenAI, Anthropic, Google), comparing subscription plans and pay-per-token services for cost, permitted use, and data handling; recommend and configure the selected approach.
- Select and configure a coding harness (Claude Code, Codex, or Antigravity), shared skills, agent instructions, and deterministic controls; train the pilot team.
- Implement CI/CD in GitHub Actions or Azure DevOps, including automated tests, review gates, isolated parallel work, approved production promotion, and rollback.
Done looks like: Evergreen has a governed agentic engineering workflow and CI/CD foundation in place, with the team ready to use them.
Phase 2 — Data Platform & Ingestion
Duration: one month
What we will do
- Compare Azure SQL and Microsoft Fabric on cost, scalability, and operating requirements; recommend and implement the selected platform.
- Configure access, storage, backup, and recovery; deliver platform configuration through CI/CD.
- Implement scheduled ingestion from agreed Workday and legacy/archive sources, covering up to ten source entities or endpoints agreed at phase start, with validation, restart handling, and failure alerts.
Done looks like: A modern data platform is in place, with priority data sources connected and a foundation for continued expansion.
Phase 3 — Agentic-First Reporting
Duration: one month
What we will do
- Compare Microsoft Fabric/Power BI with an open-source approach such as DuckDB and Cube, including licensing, hosting, and support costs; implement the selected environment.
- Build a governed semantic layer with shared metrics and controlled agent access.
- Deliver the priority reports agreed at phase start and train Evergreen's team to maintain and extend them.
Done looks like: A reporting environment is in place, with initial reporting available and the team equipped to expand it using agentic workflows.