Learning curve while we build our first Google Assistant enterprise solution
First thing first. After developing some common and simple GoogleAssistant & Alexa apps, we build our first enterprise solution for Bombay Stock Exchange (BSE). BSE is Asia’s first stock exchange and now world’s fastest stock exchange with speed of 6 microsecond. The app was officially launched by Secretary, Department of Economic Affairs, Ministry of Finance, Government of India. Within one month our beta launch active users surged to 1500+. Average conversation length is 100 seconds which implies engaging conversation with the user.
So let’s come to the learning curve/challenges we faced while developing the solution.
- Webhook – Since the solution was for a stock exchange, data security was the top most on the list. The challenge was around setting up a dedicated server and linking it to Dialogflow through webhook.
- Getting reserve word BSE approved – This was one fell of a task. Though we followed all the process as highlighted by AoG, it somehow didn’t work. I would highlight Google Assistant support team were really co-operative. They were prompt in replying to our queries and also guided us for the work around to link the reserved word to our skill.
- Understanding user flow – Though the users of the app would primarily be stock market investors, understanding the user journey was important. We did not want user to continue with a pre-defined flow, therefore we hardly used slot filing. Slot filing binds the user (without which they cannot move forward) and changing course of the conversation by the user becomes difficult.
- Provide contextual responses – User – “Tell me the price of Infosys”, BSE – “Infosys is trading at xxxx”, User – “Reliance Industries”, BSE – “Reliance Industries is trading at xxxx”. Here, though the user didn’t specify for the stock price of Reliance, the app understood it was looking for stock price as the context has been same.
- Using contextual Fallback – A user might ask for information on a company that is not listed on BSE. In this case the Fallback response would be something like – “This company is not listed on BSE. Tell me a valid company name”, while for other fallback the response would be different. Building a contextual fallback around different scenarios was indeed quite challenging.
- Logic building around the responses – Response were build around a) user’s earlier interaction with the app b) Making calculations are backend and showing the most appropriate response c) Random multiple responses for each query.
We @ITMINES believe voice as an interface is the next big change to happen in consumer behaviour. We are doing some serious stuff in BFSI, Retail and consumer segment for large enterprises. You can have a quick view on few of our demos here – http://www.itmines.com/voice/ . Happy to share more such experiences as we move along to build top class voice solutions for our clients. Till then, Cheers!!!
IT Mines Technology Pvt Ltd