We started to build chatbot two years ago when Telegram was announced their platform and then other messengers do the same step.
In the beginning, we have created a Converter Bot, it converts any value from one metrics to another
– Telegram: https://t.me/Converter_robot
– we also created this bot for Skype and Slack, but we don’t support them now when MS change requirements
One of our ideas was a bot for local cinema https://t.me/portalcinemabot for telegram messenger. It is simple but popular and provides information quickly
These were simple bots to check the API of each messenger.
Then we have created several bots for companies which want to make an interactive FAQ when a customer asked a question in the native language and get responses. For such bots, we have used the NLP services. We have checked LUIS.ai, WIT.ai, dialogflow.com. In the beginning, we have used LUIS.ai. It back multiple variants with the different value of the accuracy and we use it to ask customer additional questions and teach our bot to better understand the customer. It was the bot for https://cpf.gov.sg/ and help customers to choose the pension.
When we have worked on the bot for Acorns we have used dialogflow as NLP backend https://www.facebook.com/
In the last case, it was the chatbot that used for safety monitoring and detected the people without a helmet and safety vest and sends alerts to the manager to the chat. Out systems was showing on the https://www.
We have written several posts in our blog about our experience with the chatbot development:
– https://onix-systems.com/blog/
Please, ask any questions if you have.