An AI prompt is the instruction you give a language model to tell it what to produce. In a spreadsheet context, the prompt is a template you write once and apply to every row, with each row's column values substituted in before the model runs. For example, "Write a one-line opener for {{Company}}" becomes a different concrete instruction for every row in the sheet.
Why prompts matter for row-level work
The quality of the output is bounded by the prompt and the data it references. A prompt that only sees a company name produces generic text. A prompt that references scraped facts or verified fields produces output specific to one prospect. This is why prompts in ReplyLabs reference columns by header rather than hard-coding values.
Parts of a good row-level prompt
- An instruction. What to produce and in what format, for example "one sentence, no greeting."
- Column references. Placeholders like
{{Company}}or{{Notes}}that pull each row's real values. - A fallback rule. What to do when an input is missing, so the model never invents facts.