The Executive Deck That Builds Itself
Client: the technology group of a national facilities operator (1,000+ locations). Details anonymized; specifics available on a call.
The problem
Every two weeks, the engineering team presented a sprint-review deck to leadership: support-ticket trends, issue breakdowns, transaction volumes, and store-count changes across four product areas. The data lived in the company data warehouse and Jira — but assembling the deck was a human job. Several analysts spent the better part of an afternoon, every sprint, running queries, pasting numbers into PowerPoint, restyling charts by hand, and writing talking points. The deck was late if people were busy, inconsistent when they were rushed, and occasionally just wrong when a number got fat-fingered in transit.
What I built
A pipeline that treats the deck as a build artifact:
- Locate and fetch the upcoming deck from SharePoint automatically.
- Query the warehouse for every metric — current sprint, prior sprint, deltas, top issue categories — through a governed, read-only access layer.
- Rewrite the deck in place: highlight text, natively themed charts (brand palette, correct number formats), and rolling chart windows that shift forward each sprint.
- Draft the speaker notes — a written walkthrough of what changed and why it matters, generated from the same numbers on the slides, so the narrative can never disagree with the charts.
- Stop and wait. Nothing publishes until a human reviews the finished deck and approves. The pipeline does the typing; the presenter keeps the judgment.
Results
- Deck assembly went from an afternoon of manual work to a ~10-minute review, every sprint.
- Numbers come straight from the warehouse — no copy-paste layer where transcription errors used to live.
- Styling is identical every time, because it's code, not memory.
- The same pattern was then reused for a second recurring review — the marginal cost of automating the next report dropped to a fraction of the first.