The prompt I use for variance commentary

Variance commentary is one of the highest-volume writing tasks in finance. Every month. Every quarter. Every forecast cycle. Every function, every region, every P&L line.

It is also one of the easiest tasks to accelerate with AI, if the prompt is built correctly.

Most finance teams prompt AI for variance commentary the generic way. "Write a variance commentary for the following numbers." The output is directionally fine and operationally useless. Too generic. No structure. No tone. Often inaccurate on the specific drivers, because the model is guessing.

The prompt that actually works has four parts.

Part one is context. Not just the numbers. The business context that the reader will need to understand the numbers. What the quarter was supposed to look like. What the known drivers were going in. What the company has told the board or the market to expect.

Part two is the data, structured. Not just a line item table. The actuals, the comparison points, and the drivers for each meaningful variance, laid out in a format the model can read cleanly. Garbage in produces garbage out, and most variance prompts fail at this step alone.

Part three is the output specification. What format. What tone. What length. Who the reader is. Is this for the board, the CEO, the audit committee, or the operating team? Each of those audiences needs a different register, and the prompt has to name which one.

Part four is the constraints. What the model is not allowed to do. No invented drivers. No speculative commentary beyond what the data supports. Every claim has to be traceable back to a number in the input. This is the step that makes the output defensible.

A variance prompt built this way produces output that a finance leader can edit in ten minutes instead of write from scratch in ninety.

The variance prompt is one of the fifteen in the toolkit I built. Each one is engineered the same way, for the specific moment where a finance leader has to ship an executive-quality output. Link in comments.

#YourCFOMindset #AIinFinance #FPandA