RMEP – Financial Reporting 2.0: Automated Portfolio Reporting

Starting point

The monthly portfolio reporting for the investment committee and shareholders ran through a manually maintained Excel process: time-consuming, person-dependent and hard to scale. The data from ten portfolio companies was spread across Dropbox Excel files and a complex cash flow model — every reporting cycle hinged on individual people as a bottleneck.

Solution

Building an agentic end-to-end pipeline based on Claude Skills: automatic data retrieval of the PortCo Excel files, completeness and plausibility checks, deterministic calculation tools in Python, automatic generation of the multi-page report PDF, and a hard QA gate that validates the calculated data (aggregate = sum of the parts, P&L bridge, comparability rules). GitHub Actions serves as the serverless workflow runner, Claude Code CLI as the agentic execution layer, Slack/email as the human-in-the-loop interface — including a self-repair mechanism via automatic pull requests.

Technology stack

• Claude Code CLI + Agent Skills (agentic execution layer / report generation)
• GitHub Actions (serverless workflow runner, scheduling, versioning)
• Python (deterministic calculation and QA tools)
• Dropbox API (data retrieval of the PortCo Excel files)
• Google Sheets API (central master data / single source of truth)
• GitHub Pull Requests (controlled self-repair mechanism)

Result & Impact

• Monthly portfolio report with no manual creation effort — previously several person-days per month
• A person-independent, versioned and traceable process instead of an Excel bottleneck
• Deterministic QA gates ensure data quality before every delivery
• Blueprint for further automation use cases at the client (from data architecture all the way to BigQuery & Chat-with-your-Data)