Skip to Content
AI AgentTalk to the Agent

Talk to the Agent

Gridie’s AI agent is an AI colleague that directly understands and operates spreadsheets. Type what you need in natural language in the chat window, and the agent will read the current workbook’s data and carry out the task.

Agent chat panel overview

Getting Started with the Agent

When you open a workbook, the chat input appears automatically on the right side of the screen. Type what you want to do and press Enter or click the send button.

Agent input location

The agent automatically reads the content of the workbook and sheet currently open, including the range you select with your mouse. You don’t need to say things like “look at the data in Sheet1.”

Response Mode and Model

Below the input on the left, you can choose the response mode and which AI model to use. Once set, your choice carries over to your next conversations.

ModeWhat it doesBest for
FastOptimized for quick responsesSimple cleanup/transforms, short questions
DefaultBalances speed and quality (default)Most everyday tasks
MaxDeepest reasoningComplex analysis, multi-step tasks

Credit usage is the same across all three modes, so pick Fast for quick turnarounds on simple work and Max for tougher analysis that deserves more care.

From the same menu, you can also choose the AI model between GPT (OpenAI) and Claude (Anthropic). When a result isn’t what you expected, switching models sometimes helps.

What Can You Request?

The agent can handle almost any spreadsheet-related task.

Data Cleaning · Transformation

  • “Remove all spaces in column A”
  • “Standardize the date format in column B to YYYY-MM-DD”
  • “Find and remove duplicate rows”
  • “Merge the columns with first name and last name”

Analysis · Calculation

  • “Calculate the total and average sales”
  • “Summarize sales by region”
  • “Pick only the top 10 items”
  • “Calculate the month-over-month change rate”

Formatting · Visualization

  • “Highlight rows where sales exceed $10,000 in yellow”
  • “Create a bar chart from this data”
  • “Make the header row bold with a blue background”

Sheet Structure Changes

  • “Create a new sheet and name it ‘Summary’”
  • “Copy the range A1:D10 to a new sheet”
  • “Create a pivot table showing monthly totals”

Understands Complex Sheets on Its Own

When working on sheets with large amounts of data or complex structures, the agent carefully examines the sheet before starting. It identifies header positions, column meanings, and data ranges before proceeding, which increases the likelihood of getting the results you want without having to explain everything.

This is especially useful when referencing data from other sheets or requesting tasks that span multiple sheets.

"Analyze the data in Sheet2 and update the summary table in Sheet1" "Understand the structure of the raw data sheet and create an aggregation sheet"

Specifying Cell Ranges

When targeting specific cells or ranges, you can mention the cell address directly or select a range with your mouse before making a request.

"Sum all numbers in the range A1:C20" "Clear the entire column D in Sheet2" "Bold the currently selected area"

If a sheet name contains spaces, enter it exactly as it appears — for a sheet called Sales Data, type Sales Data, spaces and all.

Reading Data from Files and Images

Attach image or PDF files to the chat window, and the agent will read the file contents and extract data or summarize the information. Scanned receipts, PDF invoices, photos of tables, and more can be entered straight into the spreadsheet.

"Read the date, items, and amounts from the attached receipt and enter them in the table" "Extract only the numerical data from this PDF report and organize it" "Copy the table in the photo into the sheet as-is"

Supported formats: images (JPG, PNG, WebP, etc.), PDF, Excel (xlsx), CSV, text/Markdown, DOCX, HWP, and more.

To turn an entire Excel/CSV file into a new workbook, use New (+) → Upload workbook in the sidebar instead of attaching it to the chat.

Data extraction from PDF example

Searches the Web for Information

When you need up-to-date information or data not in your sheet, the agent can search the web directly.

"Find today's exchange rate (USD/EUR) and enter it in B1" "Search for Apple's latest quarterly earnings and summarize them" "Create a comparison table of current subscription plans for three competitors"

For structured data like stock prices, financial statements, or payment history, connectors retrieve it more accurately than a web search.

Fetches External Service Data Directly

The agent can directly query data from connected services (connectors) like Stripe, DART, Google Analytics, and BigQuery — pulling payment history, financial statements, or visitor statistics straight into your sheet.

"Get this month's Stripe payment count and total revenue" "Find Samsung Electronics' operating profit for the past 3 years from DART" "Show last week's daily visitor count for our site"

Connectors are exposed as agent tools only when activated. Turn one on by asking in chat (“turn on the OO connector”) or connect it ahead of time, and add your own tools by connecting a standard MCP server. See the connector docs for the full list of available services and how to authenticate.

Connectors & Credentials · Custom Connectors (MCP)

Changes Table Styles All at Once

After organizing data, the agent can also make the table look polished. Instead of specifying color codes or options one by one, just describe the style that fits the purpose.

StyleCharacteristicsBest for
cleanBlue header, striped backgroundGeneral business reports
compactSmall font, high-density layoutViewing lots of data at a glance
financeDark navy header, automatic comma formatting for numbersFinancial statements, settlement documents
minimalNo background color, thin divider lines onlyClean presentation materials
"Apply the finance style to the range A1:F50" "Change this table to minimal style and freeze the header row" "Apply the clean style and auto-fit column widths"
Table style preset examples

Creates Charts and Pivot Tables

You can request data visualization verbally. Specify the chart type, range, and title all at once, or modify an existing chart.

"Create a line chart showing monthly trends using the sales data in column B" "Create a bar chart comparing team performance and title it 'Q3 Team Results'" "Create a pivot table showing this data by department and month"

Automatically Colors Cells Based on Conditions

You can also set up conditional formatting that automatically highlights cells meeting specific conditions. Once set, it applies automatically whenever the data changes.

"Highlight cells where sales are below the target in red" "Emphasize the top 10% values in green" "Color entire rows where the deadline is before today in gray"

Works with Tables and Pages Too

The agent doesn’t just work on spreadsheets — it can also create and edit tables and pages.

  • Tables: Datasets you query with SQL for large volumes of data. Ask in natural language and the agent inspects the schema, writes SQL to query or aggregate, adds new rows, and updates the table.
  • Pages: Document- or dashboard-style pages. The agent can weave analysis results into text, tables, and charts to build a report page.
"Use SQL to get the total sales by region from this table" "Organize the analysis results into a monthly report page"

Tables · Pages

Continues Work Across the Conversation

The agent remembers previous results within the conversation flow. You can use intermediate results directly in subsequent requests.

"Calculate total sales for 2023" → "It's $1,234,000." "Use that value to calculate the achievement rate against the target" → Agent uses the just-calculated value in the next step

If you have a flow you use repeatedly, save it as a workflow. Next time it’s done with a single run — and with a schedule, it runs on its own at the time you set.

Create a workflow · Schedule

Viewing and Restoring Chat History

Your conversation with the agent is saved automatically. To review previous chats, click the Chat List button at the top of the chat panel.

Chat list screen

Selecting a previous chat from the list restores its context. You can also delete chats from this list.

When the Agent Makes a Mistake

If the agent did something other than what you intended, use the “Revert to this point” button shown between messages in the chat history to restore the workbook to that point. Then add more specific instructions and try again.

The more specific you are about sheet names, column names, and conditions, the more accurate the results — and the less you’ll need to revert.

Version History

Next Steps

  • Save repetitive tasks like a function → Workflow
  • Run workflows and messages automatically at set times → Schedule
Last updated on