Getting Started with SAS Viya Copilot for SAS Visual Analytics

The SAS Viya Stable 2026.04 release introduces the SAS Viya Copilot. In this post, I will cover how SAS Visual Analytics leverages copilot and walk through its capabilities and interface, best practices for writing prompts, as well as some tips on responsible use.

Select any image to see a larger version.
Mobile users: To view the images, select the "Full" version at the bottom of the page.

Capabilities

Graph Summaries

If you are sometimes overwhelmed in interpreting a visual, this feature might help. For selected graphs, you can right-click on the graph to generate a natural language graph summary that explains key insights:

The list of supported graphs keeps expanding, check SAS Help Center: Generate an Object Summary (select your Viya version at the top) for to an up-to-date list.

Report Localization

If your report viewers speak different languages, you can localize (translate) the labels, tooltips, and other descriptive text that is part of your report. Note that this does not translate the values of category data items (although you can also ask the copilot to create a translated version of a category). To localize a report, click on the report menu and Localize Report. Previously, you had to translate each label manually. Now, you can easily generate these translations. Click on New translation and select a language. Then, you can optionally click on the Settings button to provide some context to improve the translation. Then click on Generate. Then, review the individual translations and click on OK. Voilà, the localized version of the report will be displayed if the report viewer configured their browser to use the localized language.

For more information, see SAS Help Center: Localizing Reports.

Create custom categories

Custom categories allow you to recode category and measure data items. For categories with many distinct values, recoding can take some time. Now, you can use the generative AI capabilities of the copilot to generate value groups and assign values to those groups. Note there is a cardinality limit (1000 values) for this feature.

Say for instance, you have order data classified by country but instead you want to see the data by Hemisphere (with the values Northern and Southern). In the Data pane on the left, click on + New data item and Custom Category. Then, you can give it a name and specify which data item the custom category should be based on. Optionally, you can start creating the value groups yourself.

I start creating a group called Southern and place Singapore in it because, let's say, there is a business reason to group it with the Southern hemisphere countries. This makes sense given its close proximity to the equator (even though it is North of the equator). Now, I can click on the Settings button, and provide context to help create the value groups. Note that when the copilot generates new value groups, it keeps the value groups that were manually created (in this case Southern). Furthermore, we can ask it to keep the values assigned to those value groups as well (in this case Singapore) by unchecking Replace pre-existing group values. Finally, click on Generate value groups and review and adapt the groups as necessary.

For more information, see SAS Help Center: Working with Custom Categories in a Report.

Chat interface

As of Stable 2026.04, you can chat with copilot to:

find and add data sources using natural language
update a data item's properties (for example, the label, the format, the aggregation, or the classification)
create or modify data items (for example, hierarchies, custom categories, or calculated items)
create new report pages and add graphs, tables, and other content to the page
refine existing pages by adding, changing, moving, or deleting objects or other content
create filters, ranks, categorical display rules, and expression-based display rules
query your data and discover insights
query SAS Visual Analytics documentation

Refer to the About SAS Viya Copilot for SAS Visual Analytics for the latest as new capabilities are added.

To access the chat interface of the copilot, open any report in SAS Visual Analytics, ensure you are in edit mode and then click on the Show SAS Viya Copilot side pane button in the top right of the application bar.

Note the Attach button at the bottom right of the side pane.

You can attach files from your local computer or SAS Content to:

Give copilot a wireframe for a dashboard that you'd like to build.
Give copilot a color theme to use in the report. Let's say you've built your report, but would like to change the colors to match the company logo. You could attach an image of the company logo. Then you could ask the copilot to extract the colors from the image, and use those colors in the report.

Note at the top of the side pane , there is a New conversation button. When you have a conversation with the copilot, it uses the current conversation as context so you do not have to repeat yourself. For example, if you asked it to create a bar chart, and then asked it to change the orientation to horizontal, it would understand that you are talking the bar chart. When you start a new task and the previous conversation is no longer relevant, you can click on this button.

Best Practices for writing Prompts

The output you get chatting with the copilot is only as good as the prompts you write. In this section, I will talk about some best practices for writing prompts.

Referencing Objects

When asking the copilot to work on an existing object in the report, you can refer to the object by its object type (e.g. the bar chart). However, if there are multiple objects of the same type, you can use the object name. Each object is named automatically when created. You can change the name of the object using the Options pane on the right. You can also see the names of all objects in the report from the Outline pane on the left:

Referencing Data Items

If you have data items whose names contain special characters, it is a best practice to enclose the name in single or double quotation marks, followed by an n at the end. For those of you that program in SAS, this convention will be familiar - it is called a name literal. For example, Profit (1M Difference) should be written as "Profit (1M Difference)"n.

Numbering subtasks

If you have a list of tasks you want to mention in the same prompt, you can number each subtask. This eliminates any ambiguity about what constitutes a distinct step and thereby avoids something being overlooked. Numbering the tasks also ensures that they are executed in the specified order. For example:

(1) Create a new data item that calculates the number of days between "Order Date"n and "Delivery Date"n. (2) Ensure it is called "Days to Delivery"n. (3) Use a format with zero decimals.

Note here the second step (naming the data item) does not depend on the first step (we will address that case in the next section).

Chain of Thought Prompting

If your prompt contains multiple steps where the output for each step affects the next step, you can use Chain of Thought prompting. For example, consider this prompt requesting a three-month profit difference calculation:

Create a new data item that calculates the difference of Profit compared to three months ago, based on "Order Date"n.

Copilot first has to decide which function to use for a difference from previous period calculation, and then which parameter within the function to change so it computes the difference compared to three months ago. These are not simply sequential tasks as the reasoning behind step 1 (the choice of the function) is relevant to step 2 (the choice of a parameter to change in the function, as different functions have different parameters).

In these cases, it can help to break the prompt down into subtasks (as above), and to ask the copilot to provide reasoning for each subtask. In this case:

Create a new data item that calculates the difference of Profit compared to last month based on "Order Date"n.

Here, we are creating a simplified version of the calculation. Then we ask:

(1) Which function did you use? (2) Which parameter do I need to change to compare to 3 months ago?

Lastly, if necessary:

Apply this change.

Because you ask the copilot to provide reasoning with each subtask, Chain of Thought prompting can improve results.

Responsible Use

Validating Results

You remain responsible for the contents of the report, even when you use the SAS Viya Copilot. Remember that large language models (LLMs) as a technology are known to hallucinate, and this is not unique to SAS and can stem from a number of reasons. It is therefore vital that you validate all of the copilot's outputs. For example, when you generate automatic translations, go through each translation and ensure it is accurate. When you create a new data item, test it to check whether the results are as expected. If the copilot is employed in a higher risk-use case, ensure you are complying with any AI legislation you might be subject to.

Last but not least, generative AI currently uses a lot of water and energy in data centers. Therefore, try performing simple tasks like moving objects and renaming data items yourself once you know how to.

Conclusion

As you can see, the SAS Viya Copilot can assist you to accomplish tasks more easily and increase your productivity. Keep your eyes peeled for any new capabilities in this space - things are evolving quickly. For example, the copilot will soon be added to other products in SAS Viya - such as SAS Studio!