A git-backed automation pipeline that replaced an agency’s manual report-publishing and task-creation process with a single approval-gated workflow.

Brief

A digital agency produced regular client reports with the help of AI drafting, but every report still had to be manually copy-pasted into the client’s website and the action items hand-typed into a project tool. The process was slow, easy to get wrong, and left no record of who changed what or how to roll a mistake back.

Approach

I designed a pipeline built around a single source of truth: structured documents (markdown with a defined data schema) held in a version-controlled repository. When a document is approved, an automation layer reads it, converts it into a publish-ready web page, and routes any action items into the project tool as tasks — automatically. I split the automation into three clean stages — ingest, publish, and operations — rather than one tangled flow, and built in a draft → review → approved gate so nothing reaches a client until a person has signed it off.

Outcome

Manual publishing and task entry were eliminated. Every change is now versioned, auditable, and reversible, and the approval gate means nothing goes out unreviewed. The same pattern works for any document-driven reporting process, not just this one client.

My role: information architecture, the document/data schema, the three-stage workflow design, and the approval model.

Stack: Markdown source-of-truth · version control (Git) · N8N automation · WordPress · project/ops tool integration.