Chatbot for Answers to Questions
https://www.youtube.com/watch?v=gX2Q83C1qh4
Task number 1 for chatbots is to answer FAQs.
To support customers and to attract new ones it is necessary to answer questions. Lately many online shoppers prefer to buy goods through messengers. Thus, Endurance has increased the number of orders by 30% since we started to take orders through Whatsapp, Viber, Telegram, Skype, Hangouts, Facebook messenger, Twitter, WeChat, Google Allo and Line.
Potential customers enthusiastically ask questions and get answers in messengers, after which are ready to place an order and pay for it in the same messenger. It opens wide perspectives for the Internet trading and at the same time creates certain difficulties.
One of them is answering questions 24/7. This is not a big problem for one person if the number of customers is small. But at 50 or more requests a day an automatic solution is a need.
Configuration and training of a chabot
Happily, such a solution exists. This is an artificial intelligence named chatbot. It can answer customer’s questions in any messenger.
Naturally, it is difficult to train a chatbot to give correct answers to all the question. But it is not a big deal to train it to answer FAQs. In most cases customers want to know the same things: basic terms of acquiring this or that item or service and get assured they are making the right choice.
Endurance has created a few chatbots to assist customers. One of them answers questions about lasers. To train it we used the knowledge base with questions and answers accumulated during three years of the company working experience. The accuracy of the chatbot’s answers is 80-85%.
http://old.endurancerobots.com/zadat-vopros-chatbotu/
This is what you see on the website:
Another popular chatbot can tell everything it knows about chatbots.
http://old.endurancerobots.com/chatbot-o-chatbotax/
It tells about present solutions on the market, the price of the chatbot development, about its benefits and advantages, compares the features of various platforms.
Armed with this information users can decide if this subject is of any interest and if they are willing to go further and discuss a business case.
The most recent Endurance solution is a chatbot telling about the ICO Blockchain project INS.World. As far as many ICO projects have the majority of the community in groups of such messengers as Telegram, users are comfortable to write and get answers from the chatbot in the same messenger or in the group itself. http://endurancerobots.com/chatbots-for-ico-blockchain-projects/
This chatbot is also trained to answer general questions about blockchain, crypto currency, distributed database, etc.
The projects is implemented in English.
We have a separate business case on training a chatbot to answer questions about business procedures. The task here is rather wide and complicated. First of all, there are parts in the procedures that can be omitted when training the chatbot. Secondly, there are no FAQs, that’s why the trainer should select model questions or teach the system and carry on итерационное тестирование of the answers accuracy.
http://old.endurancerobots.com/arxitektura-produkta-dlya-vnutrennego-xranilishha-bazy-znanij-enterprise-reshenie/
It should be noted that to get excellent results it is absolutely necessary to train a bot using only competent questions and competent answers. Answers must not contain private information or references to the previous dialogs.
Besides, before putting the chatbot into operation it is necessary to carry out a number of pilot tests, that is to let users or company workers check its work by means of asking questions.
And the last thing but not the least, when the chatbot is ready and operates well it needs constant improving and updating, with this in mind you check the logs and correct the answers. You do it to make it easier for the user to get full relevant answers without asking for details. Sometimes the user asks something meaning something different and doesn’t get the answer he expects. That will be a revolutionary breakthrough in the technology when chatbots learn to understand what the user actually means.
At the end I’d like to point out that so far neither users nor companies feel need in chatbots. It’s because the technology is not yet widely known. However, that will win who will start using it in their busyness earlier.
As I have already mentioned, Endurance laser sales increased by 30% after we had started taking orders in messengers, while our competitors still go on contacting their customers via phone or email.
What are chatbots? What do you need them for? – Here’s a 10 minute video story:
How to create one’s own chatbot in 45 minutes