Using the AI Analyst ✨

AI Analyst Guide ✨

The AI Analyst is a spreadsheet copilot that helps you build better spreadsheets, faster.

How to use it

Using the AI Analyst is straightforward:

1. Find the sparkle ✨ icon in the top right corner of any table.
2. Click to open the AI Analyst panel on the right side of your screen.
3. Begin your interaction by typing your questions or requests.

The AI Analyst is very versatile. To make the most of the AI Analyst, it's worth understanding its thought process and how it handles your questions.

How it thinks

Three main principles to keep in mind when asking questions to the Analyst:

1. Context: The Analyst derives its knowledge about the dataset from a sample of rows and the header. This means it's able to understand and infer operations based on the context. For example, if your table has cost and revenue columns, you can ask it to compute profit.
2. Single-table analysis: The Analyst only analyzes data from one specific table at a time. This means that even when your question generates a new table, your subsequent queries will still refer to the original table. If you want to create a Chart of this new table, you'll need to select the AI Analyst of that table and ask it there.

Future Developments

We're continually working to improve the AI Analyst. Here's a preview of what's coming:

• Multi-table context: The AI Analyst will be aware of other tables in your spreadsheet, and able to perform actions across different tables in your spreadsheet, e.g. perform a VLOOKUP across two datasets.
• New capabilities: The new versions will support new operations such as applying Conditional Formatting, removing columns, and doing Python requests.

Get the best results!

The AI works best when:

• your Tables name has a name that is easy to understand.
• same for Table headers, they are very clear about what's there. (Shorter is better too).
• There is at least one full row of data below the headers. (The header is the first row or column).
• For Tables with date/time headers (e.g., P&L reports), the AI will assume a vertical orientation and treat the first column as headers.

Now let's dive into the different capabilities of the AI Analyst✨.

1. Data Analysis

The AI Analyst efficiently processes numerical data for metrics, data slicing, and table transformations.

Good questions, ineffective questions, and tips:

• ✅ Analysis: "What is the average and median revenue for each product category?"
• ✅ Analysis: "What is our top-selling product in each region?"
• ✅ Analysis: "Provide a breakdown of our customer base by age group and average purchase value."
• ✅ Analysis: "What's the median salary for females?"
• ✅ Analysis: "How does our product performance vary across different regions and seasons?"
• ✅ Analysis: "Based on the salary data, is it worth pursuing an MBA?"
• ✅ Explainer: "What does the column 'Bayesian classifier' mean"?
• ❌ "What's in this spreadsheet?"
• ❌ "Tell me insights about our sales." Most of the times this is too generic and the answers won't be deep enough.
• ☝️ Tip: Refer to the column name you want to execute the action on. Mention explicitly the aggregation operation you want to perform or the dimension you want to use to pivot/slice the original dataset.
• ☝️ Tip: If you need to perform an operation that requires different intermediate steps, ask it one at a time. For example: on a daily sales dataset, to analyze weekly data, you need to first add a week column, and then aggregate the key metrics.
• ☝️ Tip: "What can I ask you to do?" Eventhough this is a bit of a meta-question, it can help you understand the kinds of skills Rows AI has.

Here's an example video of a user asking for a revenue summary per day.

One more video!

2. Add a Column

The AI Analyst can create new columns and understand your table's context based on column titles and sample data.

For instance, with "Revenue" and "Cost" columns, you can simply request to add a "Profit" column. The AI Analyst can also enrich your table with AI-powered functions, performing actions like sentiment analysis, data mining, or classification tasks.

Good questions, ineffective questions, and tips:

• ✅ "Calculate the profit margin for each sale."
• ✅ "Extract the numerical portion of the user_id."
• ✅ "Add a column with current stock prices for each company in our portfolio."
• ✅ "Add a 'Sentiment' column for each customer review"
• ❌ "Make a new column with sentiment analysis." Unclear what column you want to analyze.
• ✏️ Tip: Rows can add columns that are simple transformations of your data (split) or columns that fetch info from the Internet with AI.
• ✏️ Tip: To avoid misunderstanding, be clear about the logic you want to include in the column.
• 📖 We support a lot column transformations. See table below.

See the example below, where the address column is categorized into broader regions:

In the following video, instead, the dataset is enriched with the real-time stock price:

3. Web Research

Use the AI Analyst to research and scrape information from webpages in real-time. More often than not, Web Research is a special case of Column actions.

To use the researcher, make sure that the table has a column with a list of urls or webpages. Trigger it by using the keyword 'lookup' or 'scrape' in your prompt.

Examples of questions:

• ✅ "Add a column with the meta description of the pages"
• ✅ "Scrape the urls and add columns with the funding amount in millions and startup name"
• ✏️ Tip: Always mention that you want the agent to explicitly scrape/lookup the content of url and mention what data you want to extract from it.

Here's a video of the Web Research in action.

4. Add a Chart

The AI Analyst can create charts to visualize your data effectively.

Good questions, ineffective questions, and tips:

• ✅ Bar Chart: "Create a bar chart showing total sales by product category."
• ✅ Column Chart: "Generate a column chart of monthly website visitors over the past year."
• ✅ Combo Chart: "Create a combo chart with a column chart of revenue and a line chart of profit margin on the secondary axis."
• ✅ Stacked Area Chart: "Create a stacked area chart showing market share trends for our products."
• ❌ "Make a chart of the numbers." Too vague.
• ✏️ Tip: Specify your preferred chart type, and what you want to see on both axes.

Here's a video example of creating a Chart with AI.

5. Format ranges of cells

The AI Analyst can apply various formatting styles to ranges of cells, to enhance readability and highlight key information.

Good questions, ineffective questions, and tips:

• ✅ "Format the 'Total Revenue' column as USD."
• ✅ "Format the 'Growth Rate' column as percentage."
• ✅ "Format the header in bold, red background."
• ✅ "Format range A1:B10 in light blue'
• ❌ "Make the spreadsheet look nice."
• ✏️ Tip: Always mention the data range you want to format and the rule to apply.
• 📖 Formatting Options supported
  • Font styles: Bold, Italic, Underline
  • Colors: Font color, Background color
  • Number formats: Currency, Percentage, Date/Time
  • Text alignment: Left, Center, Right

And here's the video!

6. Find & Replace

You can use the Analyst to find and replace values from a table.

Good questions, ineffective questions, and tips:

• ✅ "Replace , with ; on column Addresses"
• ✅ "Remove & symbol from the Customer column"
• ✅ "Replace - with / everywhere"
• ✏️ Tip: Specific column or range you want apply the find and replace operation.

Frequently Asked Questions

What about Data Privacy in the Rows AI Analyst?

We prioritize the security of your data:

• Your data is not used to train models that could benefit other users.
• We transmit only the minimum necessary information to the AI algorithm.
• For comprehensive information about our data handling practices, please visit rows.com/privacy.

We're committed to continuously improving the AI Analyst's capabilities. Stay tuned for updates in the coming weeks and months.

How do I activate the AI Analyst in Rows?

The AI Analyst is accessible to all users. Look for the ✨ icon next to the filter and settings options of any table.

What are the usage limits for AI Analyst?

The Free plan includes limited queries per month. The Plus plan offers unlimited usage.

Which model powers the AI Analyst?

The AI Analyst uses a combination of OpenAI's GPT-4o models, Llama and other LLMs. We regularly update it to the latest available models.

Can I use a different model with the AI Analyst?

Currently, we exclusively use OpenAI's GPT-4o model. For enterprise solutions or custom requirements, please contact us at enterprise@rows.com or use the "Chat with us" option in your Rows account's support menu.

Is my data used for model training?

No. The AI Analyst does not use your data to train models that could be used by others. We minimize the data sent to the AI algorithm. For more details on our data processing, visit rows.com/privacy.

Does OpenAI use the data they receive for training?

No. OpenAI does not use customer-submitted data via their API for model training or improvement without explicit opt-in. Refer to OpenAI's consumer privacy FAQs for more information.

What data is shared with the model?

To protect your privacy, we send OpenAI a summary of your dataset, including:

• Table headers
• A sample of up to 5 rows of data
• Basic dataset statistics (e.g., min and max values)

Personal data is only shared if it's included in the first row/column or the sample rows. We do not send complete tables to OpenAI.

Is there a limit to the amount of data the AI Analyst can process?

There are no set limits. We send only a summary of your dataset to OpenAI's APIs, allowing the AI Analyst to process datasets of any size within Rows.

What should I do if the AI Analyst is taking longer than expected?

If processing time exceeds 30 seconds, we recommend waiting briefly and trying again.

Additional Resources

To further enhance your AI-powered data analysis skills, consider exploring these tutorials:

• Enhancing SEO with OpenAI
• Efficient Content Redaction
• Advanced Data Extraction Techniques
• Optimizing Keyword Research
• Customer Feedback Classification
• Implementing Sentiment Analysis
• Streamlining Data Cleanup