Baiya: an AI Personal Assistant to maximise productivity.
Role : User Research, end-to-end Product design, Usability Testing
TEA
Team : Product Manager, Backend Engineer, Front End Engineer, Product Designer, Privacy Lawyer and UI/UX Designer (me)
Mobile app | Accelerator Project | 2024 | 5 months
Overview
Baiya is 0-1 project of creating a personal AI assistant as part of a start up acceleration program with Imperial. I was working with a cross-functional team which included a Product Manager, a privacy lawyer, a backend engineer, a front end engineer and UI designer.
Problem
“What did I have to pick up at the grocery store?”
Research shows we forget over 80% of what we hear within 24 hours. Combined with the fact the we are consuming more data than ever, this leads to information overload.
52%
of people forget things on a daily basis.
76%
of people are affected by stress from information overload.
42%
of people feel overwhelmed in their day to day lives.
Interviews
Primary Research
We began with a research plan and started conducting interviews. It was a mix of 31 in depth interviews and 100 surveys where I asked how users were managing their day to day tasks, the kind of tools they used and the problems they faced. There was a common consensus that users were finding it difficult to juggle numerous aspects of their lives
Insights
Brainstorming
Workspace
I then conducted a brainstorming session where using crazy 8s we came up with over 30 different ideas ranging from a emotional guidance, tailouring to particular sector like lawyers. Using a prioritisation matrix, I decided to focus on four areas (memory issues, education, workplace and personal life
-p-1600.png)
HOW MIGHT WE leverage technology to supercharge our lives??
Competitive Analysis
What do current products look like?
Competitive analysis of existing tools showed that that they were geared towards a formal meeting, with limited time on transcriptions. There was a rise of AI wearables but these had additional expense and a steep acceptance curve.
Testing Assumptions Early
Concept Testing
CONCEPT TESTINGTesting assumptions earlyThe initial idea was to have an always on 24/7 transcription which would auto detect your voice and capture your conversations. I decided to do some concept testing to validate our assumptions and this is where we had to pivot the entire project.The term "always on" didn't make users feel comfortable and they wanted to be able to have control on when they could turn it off.
-p-2000.png)
Design Goals
Agency
​
Allowing users to consent and adjust the data along with a way to exit.
Transparency
​
Necessary information is presented in a clear concise manner.
Security
​
Making sure that user data is safe and private.
Humane
Making it relatable to the users
Information Architecture
How will this work?
​
Through numerous stakeholder meetings, the key features for the app were defined along with the different settings and functionalities.
-p-2000.png)
A/B Testing
Will users be willing to pay?
We decided to conduct usability testing with two options, a paid subscription with complete privacy and a free model with personalised adverts. Through comparitive testing with 12 participants as well as using Qualitrics where I gathered feedback around usability and general feedback .
-p-2000.png)
Transcribe
Voice to text transcription with automatic summaries, action items and transcripts.
.png)
Actionable
Create task lists, to-do items automatically from your conversations and meetings
Ask questions based on your conversations and access the crucial details
Personalised
Ai
Integrate with other apps like your calendar, grocery apps etc- User can directly add meetings or items into the basket
Third party
Integration
.png)
Privacy and Consent
​
Conversations are end-to-end encrypted and stored as text only. Users have ability to turn off recordingquiet hours and set limits to pause recording feature after 5 mins of inactivity.
Validation
Feedback from beta usersWe presented the concept of Nest to industry experts for validation and we received really positive responses including an opportunity to present our proposal to a charity working on an initiative in a similar space.
-p-2000.png)
LEARNINGS Establishing trust Through transparency, balancing between ease of use and introducing friction, this project showed how to create trust in technology. Tone of voice and microcopy Making sure the tone of voice in the app was contextual, clear but also evoked security.