TucoBot
A smart chatbot for BanyanPro
TucoBot is an AI powered chatbot which was my UX Design project for Media Specialist Practice module. I had designed an bot for a platform called BanyanPro which is a Learning Management System. BanyanPro is a platform which is being used by a lot of users around the world for learning, recruitment and hiring purposes.
The goal was to create a smart chatbot to help the users to navigate and onboard the platform and for automating service and maintenance, to replace the current manual phone support system.
Problem Statement
Since BanyanPro is a highly feature filled platform, it has a number of subpages and fields which makes it hard for the user to navigate through when they have to do something. Also with the high user base, there tends to be some minor issues when all the users take an assessment or have to configure their classes for such large groups. These issues are currently being resolved by a manual customer service representatives ove the phone, email or other mediums. This also poses a problem during holidays and the delay in the fixes of the bugs
Solution
Chatbots are used to help users with various things such as navigation, customer service, and maintenance issues. Research show that a chatbot , would reduce the cost of customer service by upto 30% for an organisation. Users also cited that the "24 hour service” as a major advantage and the one of the main reasons for using a chatbot. Hence this led to the designing of TucoBot!

Tuco's Avatar
Design Process
My design process included 5 stages over which the the foundations and the features of the chatbot were defined and modelled in such a way that the users would be encouraged to use the service.

Empathize
This is the first stage of a design process, I had conducted a number of interviews and surveys amongst various groups of users who use the BanyanPro Platform, ranging from students, to instructors, to employees. This had helped me find the right problem to solve and to bring the right product for the users. The survey results had brought a lot of new insights for me to define the scope of the chatbot.
Link to survey:
Define
After the analysis of research results, I was able to define the user base and come up with three type of user personas, and their user journeys had given me a clear visual of the usage and the flow that the users would under go if they were to use the chatbot.

Ideate
This stage included a number of various steps, ranging from architecture diagram to the use case diagrams and so on. This made me finalise the steps and the flow of the chatbot. I was also able to figure out the flow of the AI and how the NLP would be able to learn and train itself for future development to help the users interact better with the chatbot.

It started with a basic mind mapping or brainstorming session which made is easier to structure the activity diagram which is shown below.

Architecture Diagram


Flow for training NLP, NLU and NLG subsets
Usability Testing
I had conducted a usability testing for TucoBot with some of the users. I had three testers from various field to test it out. All three were given three of the major tasks that the bot would supposedly be able to simplify.


I had conducted a usability testing for TucoBot with some of the users. I had three testers from various field to test it out. All three were given three of the major tasks that the bot would supposedly be able to simplify.
The outcomes were:
-
All three users felt the bot was easy to navigate and find the options
-
They liked the bots persona which wasn’t too quirky for the users to interact with.
-
One of the users faced some issues when they were unable to access the courses, since they weren’t being redirected to the webpage (which was later fixed in the final product)
-
One user felt that the flow for assessment detail was a bit confusing
-
They were satisfied with almost all the features of the service ticket filing system and the flow in the chatbot.
High Fidelity Prototype
Using the results from the usability testing I was able to refine my high-fidelity prototype with all the features that were defined in the earlier stages.
Software used for prototyping was Botsociety.io
I was able to create a top to bottom prototype for my chatbot with the ability to export its conversations to a working chatbot using Rasa.ai for future developments.
TucoBot Prototype on BotSociety.io

