Amazon Q Walkthrough
GenAI feature enabling Amazon QuickSight users to obtain rapid data visualizations from simple, plain language queries.
Role
Technical Writer.
Timeline
3 months.
Team
1 manager, 1 mentor.
Stakeholders
Under the Amazon Quicksight Q (now Amazon Q) product team, my objective was to develop documentation for the natural language query and deep learning feature prior to its initial release in 2021.
With my experience in relational databases and internal stakeholders’ expertise in business intelligence (BI) analyst behaviors, I crafted a walkthrough to onboard Amazon QuickSight customers onto the feature.
Update: while GenAI capabilities have been rapidly expanding, my documentation still maintains its relevance as of 2025.
Problem
How might we guide users unfamiliar with AI/ML concepts to implement Amazon Q effectively?
Outcome
1 walkthrough article (Making Amazon QuickSight Q topics natural-language-friendly).
1 sample database, complete with 75k rows of synthetic data that was included with the feature’s release.
5+ future articles written based on my work.
Internal processes to utilize sample relational databases within the documentation team.
Rapid iterations to create artifacts that would support the team short and long-term.
Technical Writing
Developing my skill as a writer to appropriately scope, communicate, and edit instructions to utilize Amazon Q.
Experimentation
Continuously learning how to use Amazon Q in order to inform my writing and my understanding of the overall product.
Breadth
Utilizing my database design knowledge and visual communication skills to devise new ways of abstracting Amazon Q’s inner workings.
Crafting documentation with a constantly-iterating product.
Because Amazon Q was in development, I prioritized creating a walkthrough to explain how users could add metatags to their data, as well as simple queries that aligned with how one would write in SQL.
I also had the following goals in mind:
Enable Amazon QuickSight Q users (administrators and authors) to onboard into Q with as little time as possible.
Frame the walkthrough to minimize the technical knowledge (databases and machine learning) needed to effective use Amazon QuickSight Q.
Showcase the unique selling points (USP) of the feature: the ability to use plain language queries and produce dynamic visualizations instantly.
Experiment with Amazon QuickSight Q and communicate findings to the broader team.
Crafting a sample relational database to simulate realistic user behaviors.
Modeling a music streaming database to input data into Amazon Q that was (1) simple, (2) replicable, and (3) easily comprehensible by the product, development, and documentation team.
To navigate legal concerns and dataset ownership issues, I created a synthetic dataset with 75k rows on Mockaroo with hard-coded distributions to test with Amazon QuickSight Q.
Draft database with sample end-user questions. With Analyst Dee's user story in mind, what information would be prevalent to stakeholders if they were analyzing a music streaming service?
Conceptual relational database model for the example music streaming service to show the basic relationships between each table.
User Story
Analyst Dee is a BI Engineer responsible for creating and maintaining dashboards on Amazon QuickSight that are published to business users, who rely on the visualizations for day-to-day business decisions.
Dee understands the datasets and data models used, but needs to be able to generate personalized visualizations at a moment’s notice.
With Q enabled in Amazon QuickSight, Dee is able to quickly transfer information about the data model and allow other users to create visualizations instantly.
Looking back: additional user research and improving engagement with the product
This project challenged me to grow my skills as a writer, a strategist, and as a communicator.
While I went in as a novice technical writer, I was able to bring my technical expertise and people-centered curiosity to provide value to Amazon QuickSight Q’s end-users and to the product team.
If I had more time on this project, I would have prioritized getting in front of users to learn more about their challenges, their mental models, and how my documentation could be changed to meet their needs. Further, I would have liked to learn more about Amazon QuickSight and Amazon Q's ideal customer profiles, as documentation can serve as a crucial touchpoint for user satisfaction.
At the same time, I was challenged to learn a tool with a technical I had rarely interacted with (natural language query, deep learning) and to figure out how to write from a different perspective. Instead of being big picture (design, object-oriented programming), it was immensely difficult for me to whittle down one action into tiny, tangible steps.
Overall, I learned to not only look at the big picture for how users interact with software, but also to dig deep into their motivations and interactions with their institution (how they get something done, how an individual’s goals fit within an organization’s goals).