Deploy Oracle Analytics Cloud AI Assistant

The following steps demonstrate how you can fully develop dashboards conversationally with the AI Assistant, bypassing the need to access individual data elements or configuration settings.

With some additional time and effort, these dashboards can be developed into highly polished reports. Ultimately, the AI Assistant puts the power of AI into the hands of business analysts, helping them to deliver more meaningful insights more efficiently than ever before.

To deploy Oracle Analytics Cloud AI Assistant:

Connect to the data source:
1. Click Create on the top right of the Oracle Analytics Cloud home page.
2. Click Connection to begin creating a connection to a data source.
  Oracle Analytics Cloud is capable of connecting to all major data sources as well as generic java database connectivity (JDBC) data sources. This example connects to Oracle Autonomous AI Lakehouse.
3. Enter the connection details for the data source and then click Save.
  A message will appear to indicate a successful connection has been made. In the event of connection failure, refer to the displayed error message to identify and resolve the issue. The most common issue with connections is an incorrect password.
Create a dataset or semantic model:
1. Click Create on the Oracle Analytics Cloud home page, then click either Dataset or Semantic Model.
  The AI Assistant can be enabled for Datasets and Semantic Model Subject Areas. This guide will demonstrate with datasets.
2. Select the data source connection created in the previous step.
3. Expand the database schemas on the left side of the page, select the desired tables, then drag and drop them into the editor.
  Oracle Analytics Cloud automatically creates joins between tables based on the data within each table. Use the join diagram to modify the data structure as needed.
4. Flip through the individual tables using the tabs at the bottom of the screen to continue preparing the dataset. Oracle Analytics Cloud has an extensive set of tools for data preparation, including embedded AI capabilities to suggest data transformations and augmentations. After the dataset is prepared, click Save and return to the home page.
Enable indexing to allow the AI Assistant to map the user's request to the underlying data model:
1. Locate and right-click the dataset on the home page or in the data catalog, then select Inspect.
2. Click Search on the left side of the Inspect window and make sure that the value of Index Dataset For is set to either Assistant or Assistant and Homepage Ask.
  The indexing configuration can be set automatically by clicking Use Recommended Index Settings or it can be adjusted manually.
3. Specify synonyms to each column in the dataset as desired to provide additional terms that end users may use to identify the associated data element.
  
  For example, you could add number of deals as a synonym for a Deals column. This improves the AI Assistant's ability to deliver the expected output over a broader range of prompts
4. After configuration is complete, click Run Now, then click Save and return to the home page.
Create a workbook:
1. On the Oracle Analytics Cloud home page, click Create, then click Workbook.
2. Select the dataset or datasets to use in the workbook, click Add to Workbook.
  Multiple datasets can be blended together in one workbook, which will be demonstrated in this guide.
3. Click the Data tab on the top of the page to blend multiple datasets together. Some matches may be made automatically.
  To create matches, hover between two datasets until a line appears. Click the line and select which columns to match between the two tables. Edit existing matches by clicking the line between the tables.
4. When you are done, click the Visualize tab on the top of the page.
Use the AI Assistant:
1. On the Visualize page, click the star icon in the toolbar in the top right corner of the page.
  This will expand a panel containing Watchlists, AI Insights, and the AI Assistant.
2. Explore the AI Insights if desired.
  These insights are created by embedded AI processes within Oracle Analytics Cloud and they are unique to the data in the workbook
3. Click the Assistant tab.
4. The AI Assistant is designed for conversational interaction with the data in a workbook. These next steps refer to a dataset containing venture capital deal information as an example. Be sure to ask questions relevant to your data. For the deals dataset, we can start off by asking the assistant the following:
```
What would you analyze in these datasets to gauge venture capital interest?
```
  The assistant responds with some advice that helps guide the user's approach to analyzing the available data.
5. To start analyzing data, we can ask the AI Assistant the following:
```
What were the deal values by year?
```
  This returns a line chart with the total value of deals over time by year.
6. We can modify the previous line chart by using the following prompt:
```
Add field country code and change chart type to stacked area.
```
  This retains the context of the previous chart but introduces more detail and refines the visual element with a stacked area chart type.
7. To add this visualization to the canvas, hover over the chart and click the Add (plus sign) icon.
8. To further develop this analysis, we can ask the AI Assistant the following:
```
Can you tweak that to look at years and industry and make it a line chart again?
```
  This prompt retains total deals value, swaps the country breakdown to industry, and reverts back to a line chart. This line chart has multiple lines now though due to the breakdown by industry. Add this visualization to the canvas using the same method as the previous step.
9. The AI Assistant is capable of producing more than just charts. We can provide the following prompt:
```
Change the chart type to a language narrative.
```
  This visualization type leverages natural language generation to output a text breakdown of the data from the previous chart. Add this visualization to the canvas.
10. To create a map visualization, we can ask the AI Assistant the following:
```
Show me a map of deals per capita by country code
```
  Add the resulting map to the canvas.
11. We can rearrange the canvas layout by dragging visualizations around and resizing them.