New to the AI Analyst?
🎥 Watch the demo video
🔢 Get started with these datasets
Introduction
Welcome to the AI Analyst guide, where we'll go through how to use the new AI Analyst in Rows. The Analyst utilizes advanced AI technology to summarize, analyze and answer questions about your tables.
The AI Analyst is available in every Table of a spreadsheet. To access it, click the ✨ on the top-right corner of the Table, next to the settings.
What does the AI Analyst do?
When you open the AI Analyst you’ll see three separate sections, which correspond to its three core features:
Quick insights
These are top level insights about your table. The AI Analyst scans your dataset, understands the key variables and provides you with a summary in 3-5 sentences with relevant KPIs about those variables, calculating averages, medians, minimum, max values and more.
You can extend the use of the Quick insights to continue your analysis or report with:
- Copy text: Copies the full sentence with the insight.
- Copy formula: Copying the formula and cell references that generated the insight.
- Add as Table subtitle: It adds the text of the quick insight to the subtitle of the table under analysis.
Deep dives
This is the second level of AI analysis. The AI Analyst extracts trends, generates pivot tables, and identifies patterns in the data. You can add any of these deep dives as tables to your spreadsheet, and use it to add Charts or Embed the results somewhere else.
Once a table is inserted, you can inspect the formula that generated the table by going to the Table setting and clicking to “Edit Source”.
Once inside the edit panel, use the QUERY formula to manually update the table. The table will also automatically update if the data in the original table changes.
Finally, you can use the Analyst to Copy the QUERY formula that generated the deep dive or remove it from the Analyst hub.
Custom questions
Finally, you can use the AI Analyst to ask question about the dataset. The Analyst can answer your queries, making it much simpler to find specific trends, create pivot tables and extract specific information about the dataset.
You can ask it two types of questions:
Questions to extract single-values from the dataset
1Q: “What’s the total number of clicks for the Spring 2023 campaign?These questions will add new Quick insights in the Analyst hub.
Questions to extract tables from the dataset
1Q: “What’s the median number of likes of my posts for each week?”These questions will add new Deep dives in the Analyst hub.
Want to ask a great question?
Be specific in the insight or metric you’re looking for. Explicitly mentioning the name of headers that are part of the question will increase its accuracy.
🟢 “What’s the total number of clicks for the campaigns from the Organic channel?”
🔴 “What’s the total clicks for Organic campaigns?”
You can also use the AI Analyst more creatively. Here are a few recommendations of advanced ways to use the Analyst:
| Generate pivot tables |
|---|
| You can use the Custom Question block to explicitly ask the analyst to generate a pivot table. |
Q: "Create a Pivot table: Date in Rows, Store in Cols, Amount in value" |
| Exclude datapoints original table |
|---|
| It is often useful to cleanup your base dataset by excluding data that it not relevant to the analysis. You can ask the Analyst to recreate the table excluding all rows that meet a specific criteria |
Q: "Rebuild table excluding rows where Weekday is Saturday or Sunday" |
This assumes that the original table has a table names Weekday with the days of the week.
| Remove outliers |
|---|
| To go deeper into the analysis you'll often need to remove values that are higher/lower than a computed value from one of the variables. And example would be to exclude rows if column X with greater that 3x average. |
| You can do that by breading the questions in two steps: |
| 1. Calculating the auxiliary metric. |
Q: "What's 3x the average of Sales?" |
| Once the result is calculated, copy the value of metric from the text. |
| 2. Recreating the table excluding the outliers |
Q: "Rebuild table excluding rows where Sales is more than 15,550" |
This will generate a Deep Dive with a table that rebuilds the original table excluding the outliers.
Preparing your table for the AI Analyst
Before using the AI Analyst, make sure that:
- The table contains headers in the first row of the table.
- Each header has a unique name.
- In addition to the header, the table has at least one full row of data.
Additionally, the AI Analyst infers the orientation of the table based on its headers. If the headers on the first row are dates/times (as in a Profit & Loss report), the AI Analyst will assume that the table is vertically oriented and treat the values on the first column of the table as its de facto 'headers'.
⚠️ If the table fulfills all of these requirements and you're still not getting any insights, try re-importing it (or copy-pasting the data) to a new table and trying again.
Data privacy
Privacy is important to us. The AI Analyst✨ does not use your data to train models that could be employed by other users. We also minimize the data we send to the AI algorithm, and you can find more about how we process your data here.
The AI Analyst is still in beta. Give it some time as it learns new tricks in the next weeks and months.