From Raw Exports to a First-Draft Report

Writing the monthly or board finance report from scratch every cycle is one of the most time-consuming, least-differentiated tasks in finance work — and it's exactly where AI drafting genuinely earns its place, if you get the input-output relationship right.

The relationship that matters: AI drafts narrative from numbers you already trust — it never generates the numbers themselves. This is the same discipline from the foundations course, applied specifically to reporting. By the time AI touches your report, every figure in it should already be computed and verified — from your KPI rollup, your categorization output, your reconciliation results. AI's job is exclusively to turn those verified numbers into readable prose.

What a good first-draft prompt looks like. Vague prompts produce vague, padded reports. Specific prompts, fed verified numbers directly, produce something close to genuinely usable:

"Draft the narrative section of a monthly finance report for leadership. Here's this month's verified data: [cash runway, burn rate, AR aging summary, budget vs actual by category, any notable reconciliation findings]. Structure: one paragraph on overall cash position, one paragraph on spending vs budget, one paragraph on anything that needs leadership attention. Tone: direct, factual, no filler language. Every number you use must come directly from what I gave you — don't round differently or add context I didn't provide."

That last instruction is doing real work — it's an explicit guardrail against the exact failure mode from the foundations course: AI restating a number slightly differently, or adding a plausible-sounding detail that wasn't actually in the source data.

What a first draft is good for, and what it isn't yet. A first draft from this process is genuinely useful as a starting point — it gets you from a blank page and a pile of verified numbers to actual prose in seconds, and it's usually structurally sound. It is not yet a report you'd send. Every draft needs the verification pass this course builds toward, and every draft benefits from your own editorial voice — the parts that make it sound like your org's report, not a generic template.

Why this is a genuinely large time savings, even with the verification step included. The slow part of report writing is rarely the verification — checking ten numbers against a source you already trust takes minutes. The slow part is usually staring at a blank document deciding how to phrase "expenses were up but for a good reason." AI collapses that blank-page problem; the verification step (covered fully later in this course) stays exactly as rigorous as it should be, and the net time saved is still substantial.

▶️ Try this

Using the KPI rollup or categorization output from an earlier course's exercise (or a realistic mock version), write a report-drafting prompt following the structure above — feeding it your already-verified numbers directly. Read the resulting draft and note anything that sounds slightly different from what you actually provided, even if it's minor.