Objective
The Goal
Design an intuitive and error-free chat interface for users to easily access key metrics about their document uploads, such as accuracy percentage, processing times, and document counts.
Implement the first version as a bot-driven interaction, rather than an NLP-based interface, to minimize errors and ensure a consistent user experience.
Simplify the retrieval of critical information, enhancing both efficiency and user engagement within staple
User Stories
Things the users want to achive
I want to interact with a chat interface to get an overview of the number of documents I have uploaded
I want the chat feature to provide me with the accuracy percentage of my scanned documents
I want the chat to provide document-level and field-level accuracy
I want to know the processing time taken for each document upload via the chat
I want to select the date range for which I want to view document metrics
I want to filter the documents based on the queues and groups they were uploaded to, so that I can analyze document metrics for specific workflows or departments.
Wareframing
Wireframes
To understand the flow we started with the some wireframes
We decided to present the user with buttons so that he can arrive at the results fast without any error
To select the date range we planned to have a popup with date selelction
Designing
Flow:
Staple Chat is made as a new section in the dashboard so that it is easy for the user to find
While arriving at the staple chat chat screen the user is able to see a welcome message along with the options to choose from
These options are presented as cards for better clarity, the cards are equipped with icons and text
Designing
Flow:
The users are provided with options to choose date ranges as shown below, it is presented as teritiary buttons
For quick date range selection there is a set pf preset dates, this is decided after discussion with the sales team
The users are also give the option to select custom dates using a date range selector
Designing
Flow:
The users are provided with options to choose date ranges as shown below, it is presented as teritiary buttons
For quick date range selection there is a set pf preset dates, this is decided after discussion with the sales team
The users are also give the option to select custom dates using a date range selector
Designing
Flow:
The users has to select the model of the document, weather its invoice, passport, purchase order…
After that the group (collection of document queues) selection is done by the similar process
After going through the whole workflow the user is presented the result they asked for, here it is field level accuracy percentage
Future Requirement From the Clients
Objective
Users should have the ability to "bookmark" their chat sessions with the AI assistant, allowing them to easily access and review those conversations later on.
The bookmarked chat sessions should be stored in a user-accessible archive or history, ensuring that users can easily locate and revisit specific conversations whenever needed.
Additionally, users should have the option to manage their bookmarks, including the ability to organize them into categories.
Users should effortlessly generate executive reports tailored to their needs, selecting specific periods such as month on month, week by week, or day by day.
Users can receive reports via email or download them from the settings page.
Enable users to select and add bookmarked chat sessions to the executive report directly from the bookmarks tab.
Designing
Flow:
I got lot of inspirations from the design of Chat GPT and Claude AI while solving for this objective
When the users click "Bookmark this session" they will see a pop up which asks them the folder which they want to save the query to
The users has also the ability to create a new folder for saving the bookmarks if needed, this feature is incorporated to give the users the ability to categorize information
Designing
Generating Executive Report
The user can add the information from the bookmark to the report they can generate
Towards the chat we ask the user if they want a report to be generated
Then the user has to choose between a "Raw CSV Report" which will be shared to them via email or a comparative report.
For comparative report the follow up question will be the type "Month Wise" "Week Wise" or "Day Wise" reporting
Designing
Generating Executive Report
On selecting the relevant options the users can share their email id to get the report associated with a single query
If the want the executive report to be generated they should bookmark the session and add the information to the report and download that from the detailed settings page
Future Scope and Enhancements
Problem Faced
From a user POV the user would expect the comparative report to be delivered to them when they select relevant options for the same but they are not getting it
The issue we faced was it is taking lot of time from the backend to generate the report if the data is large
Objective
The user should be able to chat with all the documents that they have uploaded in natural language
When the user asks for the report through the feature the UI should be able to generate that then and there
Constraints
There is chance of lot of errors that can be made by the user when using natural language to get the data
The chance of user asking the right questions is less
Conclusion
Concluding the Chat Feature
The chat feature was concluded successfully the development too 1 month and 2 weeks and it was made live to the users
The company was able to make new contracts with the clients by making chat feature as the selling point
Graphing and tables in the UI along with natural language input is under brainstorming session within the stake holders
The future design was done and can be developed
