Back to projects

Kudi Bot

Designing a chatbot for financial services

Kudi screenshot

June 2016, UX designer and writer's log:

Kudi bot is one of the products offered by Kudi Inc (YC '17). Kudi Inc is a fintech company in Lagos, Nigeria.

The challenge of this project is to design a financial service, where people can transfer money through chat, buy airtime vouchers and pay their bills. It is the first of its kind in Nigeria - to offer financial service via a chatbot.

The target audience for this product are Nigerians between the ages of 16 - 45, who are conversant with apps such as whatsapp and own a bank account.

For a delicate industry such as finance service, experimenting with a budding technology like natural language processing, artificial intellgence and chatbot results in a number of constraints that ranges from technology to security concerns. How these constraints are met and tackled are further discussed in the process below.

Process

To design this solution, I went through a series of steps to get the job done:

  1. Research - Contextual Inquiry: Since our target audience are supposed to be conversant with chatapps, I used whatsapp as my primary medium of inquiry. There were two main objectives to this inquiry:
    1. How people relate to the idea of financial chatbot by asking questions such as:
      • How often do you visit banking halls?
      • Would you prefer to do your banking transactions through chats if you could?
      • If you wouldn't, why? What are your concerns?
    2. How people communicate with bankers in real life (so as to train the bot accordingly) by asking questions such as:
      • what do you say when you want a withdrawal?
      • How do you express to the cashier that you need to make a transfer?
  2. Analysis using scenarios: Based on the data gathered, I came up with a full scale scenario in which a typical user would use the chat app based on the goals they set to achieve. Hence, there were multiple scenarios:
    • Onboarding scenario
    • Money transfer scenario
    • Bill payment scenario
    • Surfing through/Getting to know bot scenario
  3. Building the Bot's Persona: Based on users persona and data gathered, we already knew what our target audience disliked about bankers and banking halls and what they would love to see more. Together with the stakeholders, we decided the bot would be:
    • Friendly
    • Feel Human - but not too human!
    • Smart - but only about things relating to finance
    • Efficient
    • Trustworthy
  4. UX writing: With the goal of having a friendly and effient bot in mind, the following strategies were employed to achieve these goals:
    • Friendly - semi-formal tone and voice and use of emojis whenever appropriate.
    • Make it feel human - There were many ways this was achieved. An example is the 'Kudi is typing' feature that slows replies for a milisecond to not appear too programmed.
    • Smart - Answers all questions the user might have concerning it's transactions and if it does not have one, it should refer/redirect the user to a 'human'
    • Efficient - using short, simple and unambigous words and sentences to communicate
    • Trustworthy - give immediate feedback and answers questions about security!
  5. Guerilla testing: Because of the small size of the team, it was flexible enough to conduct guerilla testing almost at every stage of conception with our target audience to check if our assumptions and solutions were still right.
screenshot sample

Outcomes and lessons

A major insight gained during one of our testing is the realisation that users were hesitant to add their card or account details at the beginning of their interaction with the bot. We had assumed that that should be the way the onboarding process should be, but this only resulted in drop-off. Users wanted to get to know the bot before they could commit. So we iterated to allow users to have as much small talk with the bot as they want and they onld added their card or account details when they were ready to perform a transaction.

Kudi bot is still being used by thousands of people daily to perform transactions. The bot was deployed on many messaging apps such as Telegram, Facebook, Skype, Slack, it also has a dedicated app of its own.

Next Project

Vendors.so