Prompting Sheets

The Get Started flow asks what the project is about and builds the sheet from your answers, naming columns and pre-filling tasks rather than dropping a generic template. That step separates it from the template galleries built into Sheets, and it's closer to an AI project plan generator than to a picker of pre-made layouts. The output is meant to be edited, and the vendor treats the first pass as a starting point for your own rows.

Context is controlled with tags in ordinary places. Adding [P] to a sheet name pulls that sheet into the AI's view, and putting [-P] in the first cell of a row excludes that row, which is how you keep salaries or client names out of a prompt. Industry words in sheet titles also nudge the model toward a matching tone, so a finance sheet gets finance-flavored answers. It's a simple mechanism, and it's the one worth learning first.

The add-on adds six functions: =PROMPT() for a single answer, =PROMPTLIST() for a list that spills down a column, =PROMPTTABLE() for a decoded table of rows and columns, =PROMPTIMAGE() for generated images, =PROMPTAUDIT() for page reviews, and =PROMPTFILL() for mapping an example range onto new data. Each one accepts optional data and context arguments, so the prompt can reference a URL or a block of cells. You can insert a formula as static text or as a live formula, which is the difference between a snapshot and something that keeps working.

Prompt templates can carry placeholders and conditional blocks. Typing "Business Name?" on its own defines a variable you then call as {Business Name} inside a template, so one prompt template serves every row. Conditions are supported too: {IF:Sales > 1000} writes text only when the value clears the threshold, with {ELSE:} and {END:IF} closing the block, and the comparisons cover equality, inequality, greater and less than, plus AND and OR. Saved prompts get IDs, local ones prefixed with L and global ones with G, so a long prompt can be reused as =PROMPT("P:L0001").

A sidebar chatbot answers questions about the sheet you have open and can look up specific values on request. It ships with a button that rewrites your prompt before sending it, which helps when you know the answer you want but not the wording that gets you there. Results land in the conversation, not the grid, so this is the exploratory tool rather than the production one.

The Sheet Library lists pre-built prompting sheets you can search, favorite, and clone with a Create Prompting Sheet button, which is handy when you want a known-good starting structure instead of a generated one. Separately, any prompt formula can take a website address as its data argument and read the page before answering, with cached copies of recently fetched pages to keep responses quick.