Lean Chatbot Procedure

Yu Li, our chatbot expert, reported at the Bot Days about his experiences with the lean approach for chatbot projects. Now we tell you about it.

Author: Yu Li

Chatbot was undoubtedly the hype topic of 2017! Microsoft Switzerland organized a so called «Bot Day» and we were of course there. Based on our experiences from various customer projects, from our hackathon and various presentations, we created the Lean approach. This will help you to get to your bot quickly and efficiently.

Challenges

Thanks to the cloud, modern technologies such as Natural Language Processing (NLP) are generally accessible. Nevertheless, we face the following challenges when implementing a chatbot project:

  • How can IT quickly identify business challenges and needs?
  • How does IT inspire your business to find meaningful use cases?
  • How does IT deliver the Minimum Viable Product (MVP) in the shortest possible time?
  • How can the chatbot be trained efficiently?

Lean Chatbot Procedure

To address these challenges, ipt developed the Lean Chatbot approach.

Chatbot-Vorgehensmodell-2.jpg
Process model for the development and implementation of a chatbot

Step 1: Chatbot needs analysis

For a quicker identification of the Pains and Needs of our customers, ipt designed a questionnaire. The basis for this was our many years of experience from various customer projects. One question as an example:

ipt-expert: «Is your team often under high pressure?»
Support Agent: «When invoices are delivered to our customers at the end of the month, the number of support calls increases exponentially.»

A list of questions helps us to create a structured conversation and reminds our customers of your challenges and needs.

Step 2: Chatbot Workshop

Once we know the challenges and needs, we will organize a Chatbot Workshop. Together with our customers we develop a Baby-Bot within a few hours. The goal is to find quick-win cases or so-called «low-hanging fruits».

We first convert an existing FAQ page into a FAQ bot. This works in 3 steps: Copy! Paste! Bot! Then we develop a triage bot with the simplest, most common but most important support issues. Microsoft Azure offers products like LUIS and Qna Maker to speed up the implementation.

Step 3: Chatbot Pocathon

In order to deliver the MVP with the right platform in the shortest possible time, ipt designed the innovative approach Pocathon.

Pocathon = PoC + Hackathon.

Before Pocathon, the use cases are defined together with the business. Teams of experts from various product providers (e.g. Microsoft Azure and IBM Watson) develop the MVP at the Hackathon within a few days. In the end, the teams from the various product manufacturers pitch the best solution in front of the customer.

Step 4: Chatbot Continuous Delivery

The MVP is not the end of the journey. It's just beginning! The chatbot has to deal with more customer requests and learn from its mistakes. What is the most efficient way to train the bot?

Continuous Delivery with Active Learning Feature is the answer. Thanks to this feature of Microsoft Azure, the bot can independently suggest requests with uncertain interpretation for review. The suggested user input is then regularly reviewed and the optimization/extension is delivered subsequently.

Conclusion

Do you dream of having your own bot on day one? Would you like to digitalize your customer service? ipt will be happy to support you with our Lean Chatbot approach.