Naiad, a Gemini-powered Data Storytelling Assistant for Slides
Led UX design and conducted user research to a develop a tool for students and young professionals to tell better data stories.
Project Details
Company: Google (Industry Sponsor) | Team: 3 UX Designers & Researchers
Role: UX Design Lead, Developer and Researcher | Timeline: Aug 2025 - May 2026
Tools: Figma, Qualtrics, Google AI Studio, Apps Script, Google Slides
The Challenge
Students have done the work. They now need to tell the story. But building a story takes skill and effort, most of which is never taught.
As our team geared up to build our semester-long presentations for a class, we realized we have all the data and information from the work we've done.
BUT WHY IS IT TAKING SO LONG?
Piecing together an effective story sure takes its time and a whole other level of effort. The more we thought about it and discussed it with our peers, we realized not all of us are taught the craft of storytelling with many of us doing it poorly, sometimes just winging it.
This came at a great time too just as we were gearing up to start our Capstone Project with Google whose focus was on students
Our work, therefore, aims to investigate if AI and in particular, Gemini can have an effect on students' slide creation processes and in particular, data storytelling. Can it help students create a story faster and would this be effective?
Can AI have an effect on student’s slide creation processes especially data storytelling during slide deck creation?
Therefore, we designed a tool, called Naiad which is currently designed as an Apps Script extension that can be integrated with Google Slides and uses the power of Gemini to assist students with better data storytelling.
The Value I Delivered
I led the design & development of the tool while simultaneously acting as the bridge between the research and design processes.
For the system
Seamless Integration Experience compared to other AI tools for Slides.
For Google
Built with focus on Gen Z users + Growth in Market for Educational Tools.
For the users
24%↓ Cognitive Load
12%↑ Creativity
The Solution
Naiad helps you start building your data story via tailored recommendations of narrative and charts, reducing starting friction.
Students click on the Naiad extension to upload their project brief or provide context.
Naiad gets involved in a back and forth conversation to gain enough context about building the narrative.
Naiad builds out a starting point - an outline framework that students can edit, rearrange or bulk edit the entire narrative if they changed their mind!
Once students have their content finalized, they can scan slides to get suggestions for best visualizations to represent the content on a particular slide.
The Deep Dive
Here’s how I helped hatch the solution
Framing the Challenge
Leverage AI to assist students' storytelling but offload critical thinking and decision-making to the students, who do it best
This was the most intense part of the process. Targeting the right research direction was crucial to our project as we wanted to establish the role AI played in our project.
From having informal discussions with students, teachers and even working professionals at Google to conducting intense literature reviews, we shifted our initial decision of helping students' critical thinking to providing students with an assistant leaving the critical thinking to what humans are the best at.
Therefore, our tool now lets our users be the ones taking the decisions when building out their narrative but are able to use the AI tool as a means to reach the decision. Our new research direction was as below:
Can AI have an effect on efficiency and perceived value on storytelling during slide deck creation?
But which parts of storytelling did we want to consider?
We had TIME and BUDGETARY constraints.
Therefore, we looked into literature again, identified a framework by Lee et. al and decided to focus on the "Making a Story" and "Telling a Story" parts of it since these are where students are our primary stakeholders and would be using Google Slides the most.
Designing the Solution & Branding
Students expect AI assistants to be designed and behave like the LLMs they interact with regardless of the "new" features introduced.
We then found 5 Georgia Tech students at the end of their semesters, just as they were putting together their end-of-the-semester presentations and conducted Contextual Inquiries.
We wanted to understand their typical slide building process to build a tool that truly assists them by identifying any gaps we might be overseeing. The most obvious being difficulty with starting out on a blank canvas. This also allowed us to couple some of our previous literature review, conversations with SMEs to come up with the following design requirements.
Design Requirements and User Needs
This allowed us to research competitors, brainstorm ideas and gain feedback from SMEs across multiple phases. The loudest feedback we gathered was that the assistants need to have a conversational flow as both students and SMEs pointed out.
The feedback was quite crucial to lay out the AI conversational and interaction flows.
At the same time, I also defined the tool's branding and helped define its design system. One that closely matched Google's design language but had its own twist.
I also created the logo for "Naiad", depicting the water nymph containing the letter "N" flowing like water (metaphorical to flow of a story) with the Gemini logo.
Building the tool in Apps Script allowed for more feature flexibility than "training" Gemini to behave a certain way.
Another important choice was to choose between building out an Apps Script extension compatible with Slides or proceed with training Gems by Gemini, our team decided to actually build it out as an extension since it can be more flexible with the features.
We then set out to design the high-fidelity wireframes and prototypes containing the updated AI and user flow.
Building the tool with AI
Being actively involved in framing the research question, I led the design of the tool whilst also solely developing the same.
In order to "vibe code" the tool, I suggested and strongly insisted on the need for a "Prompt" Requirements Document (as opposed to your regular Product Requirements Document). Contributing to its creation, I helped put together parts of the PRD alongside the UX Researcher on the team.
I then set out to build the tool using Google AI Studio to build an extension on Apps Script using Gemini's API.
PRD
The process
Wireframes
Google AI
Studio + Gemini
Apps Script
extension
While this was quite a process, spending some sleepless nights the result was satisfying. Sometimes, the code breaks but that is when I understood the importance of also knowing the programming language constructs yourself as the "vibe coder" to go in and make those little tweaks.
User Testing
Ongoing and due to research restrictions, this process cannot be detailed out here.
Naiad Narratives was a clear winner, with users taking lesser time to create data stories.
We conducted 52 user testing sessions with half of them creating a presentation with Naiad and the other without, each lasting about an hour and had some interesting findings through the System Usability Scale and NASA-TLX scales that we administered.
Task completion on average took ~5 mins lesser with Naiad.
They also reported a 24% lower cognitive effort than the control group.
Naiad users reported the creativity usefulness of Naiad 12% higher than regular Slides users.
Cognitive efforts were restructured with users acting more like editors although critical thinking was still maintained.
Qualitatively looking at it, users really liked the seamless integration of Naiad with Google Slides with some even wishing they could use it.
Reflection
This project taught me a lot about adding constraints and molding your research into something specific and actionable. We need to take things at a time and need to consider the reality.
