💬 Generative AI
ROLE
UX Researcher
DURATION
Apr 2023 - July 2023
TEAM
3 Team Leads • 1 Designer
Overview
Ask mode in DataChat is a generative AI driven platform to
Leverage ChatGPT to help data science
Allow users to ask insights and questions related to their datasets
Problem
ChatGPT has changed the ecosystem of data science. Previously, users could either write code or work with UIs, such as table, forms or charts to analyze data. Nowadays, however, ChatGPT has provided an opportunity to analyze your data simply by asking questions.
We have accomplished significant work on a spreadsheet-centered Grid Mode for DataChat. As the market's direction has shifted, we aim to further empower our platform in order to revolutionize data analytics with the assistance of ChatGPT.
❌ Current Ask feature is an embedded version in spreadsheet mode.
❌ The scope of questions are limited to single dataset.
❌ Only one question/answer is retained.
❌ The absence of clear tutorials or examples is a challenge for users.
Design Process
Our customer success team surveyed current DataChat users. A primary user group identified through research was people (age 20 - 35) from different scales of businesses. This user group confirmed initial assumptions about DataChat customers, but research also revealed that other problems included how to show the dataset in the conversation, the credibility of results, and the cost of asking questions on large datasets.
Based on the initial feedback from the findings, our team identified some key pain points and created personas. Since we had short time and limited resource for this project, we also heavily relied on secondary research by checking how other products work to speed the design and development. After then, we started with wireframes and prototypes, and refined the design by usability tests in an iterative approach.
User research
Persona
User journey
📍 Mapping a common user journey by create a conversation, learn from example and use the example as a template can help reveal the improvement opportunities for DataChat Ask mode.
Wireframes
Prototypes
Workflow 1
A flexible layout that seamlessly migrates the experience from ChatGPT
Workflow 2
Ease data analysis with a Generative AI support boilerplate
Usability study (TBD)
⚠️ This is an ongoing project, we haven’t conducted any usability study yet.
What I learned
ChatGPT follow a pattern on 3 different dimensions, question, answer, and conversations.
We need consider both experienced and non-experienced GPT users.
Templates are important features to easily guide users with low learning curves.
Impact
🏆 Researched different ChatGPT applications to provide profound proof leading the design.
🏆 Created personas to better empathize with users and clearly map the user's journey.
🏆 Helped launch the first Beta version of the generative AI mode.
“The new Ask mode looks fantastic to me, and I’m excited to see how users will respond to this powerful feature in our product.”
