Neha Singh - Portfolio
concept, product design, conversation design,
research — 2019
chat bot
Juniper Networks
chat greeting
chat bot

An NLP/ML based bot that provides quick help while integrated inside the application.




To create a chatbot that lives within products and answers relevant user queries.

For complicated technical products, user queries can be very frequent. Instead of forcing users to search for and open technical documentation, interactions with bot allow for friendly and quick customer support.
loading answer

I am the lead product designer, conversation designer and strategist.

I conceptualized this feature, evangelized the need, designed its flow and conversations, and have been involved throughout the implementation phase. I work closely with the Machine Learning team, writers, editors, product managers and customer support teams as I conduct research and design the flow.

Topic opening

My research with users and multiple conversations with stakeholders indicated to me that most users avoid reading for help.

However, product-level help was a necessity and while I designed a solution to improve that experience, I knew we needed to go beyond that.

That is how I began considering the use of an assistive system based on AI inside products. The idea was to drastically simplify how we help our users.

I knew this would not be an easy project, whether in terms of adoption or implementation. An NLP and ML based bot is still a new (and resource-intensive) technology. How would I balance the constraints with user needs?


My aim with the bot was to give users short snippets of relevant answers and lead them to detailed topics if required. We do this in a user-friendly and conversational manner so the expereince is immersive.

With aim and user base identified, I began considering the type of bot to use.

Study into bot technology indicated that we could either go with a rule-based logic, pure AI or a combination. A rule-based bot would have been simple to implement but would have not served user needs.
Why introduce a bot if all users could do was to select from various options endlessly?

Since a pure AI based bot would have been risky (more prone to errors due to technical limitations), our team decided on the combined approach. We could begin with pre-defined options and in doing so gently lead users into asking their questions in the right form.

I collaborated with customer support to understand how and when to fall back to live support.

Visual iterations

Here are a few key pieces I worked on as part of the design:

Bot personality
Our bot wasn't going to be funny or snappy or witty. Our users needed quick, technical answers while using a complicated product. Our bot had to be respectful of that need and be a dependable source.
I created a persona document that allowed our team to be in sync on our bot's goals.

Conversation flow
I identified the top use cases to cover and created a the conversation flowchart for the bot.

Bot design
After several iterations, I came up with the visual design and placement for the bot. Since it was to be a part of help universe inside products, this bot was integrated into it.

Conversation design
Writing the dialogs (for greetings, acknowledgments, questions, errors, prompts, goodbyes and more) such that they follow conversation and linguistic fundamentals, I also created a set of conversation and tone guidelines that would allow other writers to write the dialogs.

Gif of topic opening
course study next