AWS Business Intelligence Blog
Generative BI dashboard authoring capabilities now available in preview for Amazon QuickSight Q customers
Amazon QuickSight customers can now try generative business intelligence (BI) capabilities in preview to build visuals, build calculations, and refine visuals by using the natural language interface of Amazon QuickSight Q. For example, asking Q “show me count of orders in 2023 by city as a map” will instantly build a geographic map visualization automatically configured with the count of the field orders filtered by year 2023. Asking for “sales year over year” in the QuickSight calculation editor will promptly build a QuickSight expression that automatically includes the correct calculation function and data field for sales. Also, refining a visualization with the command “change to table, add discount, and highlight discount > 10% blue” will convert the visualization to a table, add the appropriate field for discount, and apply conditional formatting to any values with discounts greater than 10%. These generative BI capabilities accelerate dashboard building and free up valuable time for QuickSight authors.
These new capabilities are the first wave of generative BI capabilities announced at the 2023 AWS New York Summit, and build on the early AI innovation of the natural language query capability of Q, which has enabled business users to ask questions of their data without having to write SQL queries or learn a BI tool since 2020. Generative BI in QuickSight is powered by Amazon Bedrock large language models (LLMs), which securely retain data within the AWS environment.
Build visuals in seconds with QuickSight Q
To help business analysts analyze data and build dashboards faster, we have tuned the natural language capabilities of Q to focus on creating visuals by describing the desired outcome in just a few words. This allows business analysts to focus on the task at hand (for example, visualizing sales in monthly granularity for the year 2023) without having to think about what fields to pick, what filters to add, and what visual type to pick, removing a sequence of manual point-and-click steps in BI and replacing it with the natural language query.
With the new Ask Q to build a visual on the dashboard authoring interface, analysts can ask for data fields, data values, filters to use, operations to perform, and visual types to use.
For example, Q can quickly create a filtered map when prompted to “show me count of orders in 2023 by city as a map,” a line chart comparing products called contactmatcher and alchemy when asked “sales of contactmatcher vs alchemy by month,” or a forecast when asked to “forecast sales of contactmatcher by month.”
 |
 |
 |
The new visual building interface also suggests sample questions to help business analysts see what other questions they might be able to ask and uses type-ahead to help them discover specific data values. To help business analysts, Q answers vague questions like “top products” with a best-matched visualization, and explains how it chose to interpret questions. For example, Q might interpret “top products” as “total sales by product,” and will also show alternative questions in a Did you mean… section when multiple parts of data could match an author’s question. Q also uses autocomplete to help authors discover and select specific values or field names from data, while respecting row-level security rules configured in QuickSight.
 |
 |
 |
Business analysts can also easily change visual type or add forecasts to visuals before using the Add to Analysis option to add the visuals to the dashboard being created.
Build complex calculations with ease
Calculations can be the most complex and daunting part of BI training for most business analysts, requiring months or even years of experience. The new generative BI calculation editor in QuickSight allows business analysts to build QuickSight expressions by describing outcomes they want using simple English, producing complex calculations in seconds.
The new Build for me options available in the calculation editor interface automatically choose data fields to add to the calculation, producing ready-to-use expressions. For example, a prompt of “rolling 7 day average order count” can build the QuickSight expression windowAvg(count({Order ID}),[{Order Date} DESC],7,0,[]). Generating calculations from natural language saves time and effort that business analysts spend looking up reference documentation, asking colleagues, or simply in trial and error.
Refine and format visuals instantly
Creating compelling dashboards often involves hours of visual adjusting and refining by business analyst teams, altering visual properties using multiple point-and-click steps to achieve organizationally preferred presentation formats. The generative BI capabilities of QuickSight now allow business analysts to customize visuals to achieve a specific presentation using straightforward natural language prompts. Visual customizations can be specific in the Build for me prompt, and several customizations can be included in a single prompt in order to quickly complete many visual editing tasks.
The following customizations are supported during this preview, with more coming in the general availability release:
- Change visual type, such as “change to bar chart”
- Change axis names and table column names, such as “rename Y axis to Account Manager”
- Show or hide data zoom, such as “show data zoom”
- Add fields to visuals, or specific field wells, such as “add profit”
- Change visual sort controls, such as “sort by sales descending’”
- Apply conditional formatting, such as “make profits < 0 red”
For a full list of available customization options, refer to the documentation. This list will be updated as we add support for more operations.
Your data, at your fingertips, under your control
Generative BI capabilities automatically understand language and work on any data without specific tagging or training. You can also override Q’s default behaviors by adding business or organizational context to control how Q interprets specific words; for example, by mapping alternative names to fields as synonyms or choosing which fields are best to answer vague questions like “who,” “where,” and how many.” Your data used with QuickSight and Amazon Bedrock is not used for service improvement and not shared with third-party model providers, and is encrypted in transit and at rest.
Try the generative BI QuickSight authoring experience
This preview is available to all QuickSight accounts with the Q add-on in US East (N. Virginia) and US West (Oregon) Regions.
If you’re not currently subscribed to Q, you can get started with a free trial. To enable this preview, QuickSight authors can simply choose the Generative BI for QuickSight Authors option from the QuickSight preview manager.
 |
 |
Although AWS is not currently charging for the use of these preview features including the underlying use of Amazon Bedrock, you are responsible for fees incurred for other AWS services you use. Standard pricing will apply for your use of those AWS services.
About the authors
Zac Woodall is Principal Product Manager of AIML at Amazon QuickSight. As the product leader for AIML capabilities in Amazon QuickSight, Zac applies AI to simplify product experiences making data more useful and accessible. Zac has 24 years of experience helping create some of the world’s most used enterprise and consumer software, with stints in startups, Tableau, and Microsoft before coming to AWS.
Jose Kunnackal is Director of Product Management for Amazon QuickSight, AWS’ cloud-native, fully managed BI service. Jose started his career with Motorola, writing software for telecom and first responder systems. Later he was Director of Engineering at Trilibis Mobile, where he built a SaaS mobile web platform using AWS services. Jose is excited by the potential of cloud technologies to help customers make the most of their data.