SiplyBot, a conversational chatbot designed to reduce the total customer calls by 19%

My Role

Design and craft an end-to-end conversational chatbot flow
Research and define the reason behind high customer queries

Conversational Design; User Journey Mapping

Results

Reduction in average customer calls by 19%
Providing extra time for the CX team to focus on important concerns.

To deal with the increasing number of repetitive user queries in Siply, it was high time to introduce a chatbot to automate the common questions.

The goal here was to help the CX team be more efficient by automating the repetitive queries.

Automating conversation that unfolds as a rule-based or intent-based narrative

At this point, our CX team were responding to approximately 35 to 40 calls every day. Where 40% of the queries are repetitive in nature.

The priority here was develop a quick chatbot that can cater to the 40% of the repetative queries. And rest then can be transfered to the CX team.

We chose the intent-based chatbot

Considering the scope of the project, we developed the chatbot to be intent-based, where the flow of the converstation was controlled to cater the repetitive questions.

Neat Minimalist Comparison Instagram Post (1).jpg

Ways to present rule-based conversations

In my competitive research, I found variations in how top FinTech startups are using the rule-based conversation flow.

From which the most relevant insights were:

Highlighting FAQs on top
Keeping the tone warm and conversational

Interviewing to categorise the user queries

I reached out to our existing customers asking for their current queries which helped me understand and map them from a granular level.

This also helped me understand how each conversation formed a structure, which I will be using to create a common communication flow.

Knitting the questions to form a user flow

I tried to consolidate the idea of how the actual flow will look by mapping all the possible touchpoints — where the user can raise a query.

For clarity, watch in full screen view.

Iterations based on voice and tone

User's mindset when opening a chatbot is either negative or neutral, like curious, confused, angry, etc. I used the following tones to iterate on the communication scripts:

Empathetic
Helpful
Supportive

Presenting the final conversation draft

And finally, all the conversational pointers are content together to keep the solution clear and actionable. The flow can be monitored in 4 sections:

Welcome dialogue
Query identification
Final solution
Agent redirection