The Future of Search Is Here
An AI chatbot that helps find information and files
from your Email inbox
HSE 542 - Foundations Human Sys Engr
Spring 2024
The problem
54% of employees in a workplace spend more time in searching documents than responding to emails.
Our Team.
We are graduate students at ASU, pursuing our major in User Experience. We as a team worked on the Email Chatbot. Our main goal behind the project was to use advanced technology like Gen AI to help reduce stress in the workplace.
Solution will connect securely to users' email accounts (e.g., Gmail, Outlook) using authentication protocols
Preethi Varsha Venkatesh
Shailee Shah
Inefficiency and unproductiveness in a workplace have a significant impact on employee well-being. When employees struggle with inefficiencies, such as difficulty finding files or documents, navigating complex processes, or dealing with redundant tasks, it leads to stress and productivity losses.
Proposed solution
An AI-driven conversational interface designed to streamline the retrieval of information and documents from cloud platforms such as emails
Methodology
User surveys
Usability testing
Statistical analysis
We collected 13 responses to understand users' pain points, expectations, and requirements
Conducted a within subjects usability testing for 4 users and designed tasks based on our survey responses
We gathered data from usability testing, where the users performed same tasks with the email search vs our prototype
03
Conversation Style-Flow
A user-friendly conversational interface for interacting with the chatbot.
Email Integration
01
Solution will connect securely to users' email accounts (e.g., Gmail, Outlook) using authentication protocols
02
Information & Document
Retrieval
Allow users to search for specific emails or retrieve email attachments using natural language queries
User survey
13
Responses
Our purpose of conducting a user survey was to learn
1. User's patterns on how they search for an email
2. How severe is this problem by questions such as how often do they use email, how often do they organize their emails, how interested are they in our product
3. What kind of challenges does users face in searching for an email
4. How will they use the email chatbot and what features would they expect
Figma Prototype
We created a figma prototype based on our learnings from the survey and usability testing
Users preferred a link to the email along with the chatbot’s response
-
The UI design includes information provided by the chatbot for the query, as well as a link to email for users who wish to check further.
-
This feature adds more flexibility for users.
-
Additionally, there are 'like' and 'dislike' buttons to gather feedback on the responses and better train the AI.
Users wanted the document attached to the response and also expected an option to download attachments
-
The UI design includes an attachment of the document.
-
This feature eliminates the need for users to search through emails for information.
-
If the user wants information from the document, the chatbot will also provide a summary of its contents.
-
61.5% of survey participants were in the range of 18-25 years
-
53.8% user their email more than 10 times in a day
-
38.5% never clean out/archive their email box
-
53.9% users experience difficulty in searching for an email
-
46.2% are highly interested in our product
Quantitative Survey results
Qualitative Survey results
-
Trying to find something in a long email chain is hard
-
Finding an email/ information when I have multiple emails about the same topic or from the same sender.
-
Keyword search excluding variations of the same word
-
Too much clutter/junk mail
Statistical Analysis
Measuring the time it takes to complete each task using traditional email search versus the Email Chatbot to determine if there is a significant difference in efficiency between the two methods.
Usability testing recording
Below you can find videos from the usability testing where first the user performed a task using the traditional email search and then performed the task using the GPT builder that we have created to test the efficiency.
-
Email can only find information with exact keywords
-
Email search doesn’t detect text in image
-
Users have to remember the combination of date, sender, keywords
-
1 of 4 users used the advanced email search
-
Email search doesn’t detect ‘thirty’ to find 30 in the email
Usability testing observations
Usability testing limitations
-
User knowledge - Since the users were unfamiliar with the provided email inbox, this might have influenced the results
-
Content - The email primarily contained e-commerce and spam emails
-
Top down processing - The prompts to search were given to the participants, which were created based on the survey results
Learnings and next steps
The project helped us understand LLM, User's expectations from a chatbot and the constraints in the execution of this product.
User Greeting and Assistance Prompt
The chatbot window includes a friendly greeting ("Hi! How can I help you?"), which invites interaction and offers assistance.
Search Functionality
There are multiple search options available in the chatbot interface, such as "Find Email", "Find Information from Email", and "Find Attachment", indicating the chatbot's capability to assist with different types of email inquiries.
The chatbot provides direct responses to user inquiries within the chat window. In this case, it answered a specific question about the delivery date of an Old Navy swimsuit order.
Direct Response
The chatbot interface includes thumbs up and down icons, likely for users to rate the helpfulness of the chatbot's response
Feedback Mechanism
Branding
The survey and usability testing enforced the need of creating a trust and authenticity amongst users. Since the chatbot will be connected to their email box many users raised concerns about their privacy. We chose our brand color to be purple as purple evokes an emotion of authenticity and truthfulness.
We wanted to keep the logo simple, familiar and friendly. The elements used for the logo is the image of a mail and a magnifying glass, use of simple elements makes the logo feel familiar and approachable.
Usability testing
4
Users
Tasks
We developed a model with GPT Builder to evaluate our product's performance against traditional email search methods. Our approach involved creating a comprehensive email database which we then integrated into our Chatbot's functionality. Utilizing a within-subjects design framework, we crafted a trio of tasks and engaged a single user group to execute these tasks in two sessions: first by employing conventional email search techniques, and subsequently through the use of our innovative product.
Testing Prototype
To further understand user's expectations and know what other features would they want
Execution
To create a database for real time email content and integrate with GPT builder
Privacy Concern
Working to understand privacy issues and constraints
© 2035 by Name of Site. Created on Wix Studio.
Working to understand privacy issues and constraints