In this article, I’m going to share with your the ingredients to create Wechat NL AI response and keyword-based response chatbot. So, you can learn the difference between both, and the ingredients you need to create, for the purpose to increase efficiency to communicate with your customers from now on.
What’s Wechat Chatbot
Wechat is the largest IM + social platform in China, that is owned by Tencent Group. Wechat chatbot is a program designed to simulate a conversation with humans. Its purpose is to replace us handle repeating work, for example, to answer standard questions, engage with customers, or implement marketing communication for business goals.
Basically Wechat chatbot is similar to other social chatbots worldwide, such as messenger, Whatsapp, etc is a text-based algorithm. It can work in either keyword triggered response or NL AI response (Natural Language artificial intelligence). However, WeChat account creation is different. If you are not running a business in China or you are not Chinese residence, you need to partner with a local Chinese or an agency to create a Wechat official account.
Natural language AI chatbots attempt to self learn the labeled data fed by a human being and attempt to understand the user’s voice questions and reply automatically & accordingly. So, we need to feed labeled data to coach AI business information and conversation objective, and continuously feed huge amounts of structured or unstructured raw data to the NL AI platform. Step by step, even having some accent or typo or varied questions come up in the conversation, NL AI chatbot is able to handle as well. Thus, Chatbot NL AI platform is usually more expensive, there’re many options available on the markets, such as Tencent Chatbot cloud China, Siri developer, and Google assistant developer
Also, Wechat’s simple keyword-based response chatbot is far more common, due to the ease in development. It relies on the user only typing a set of words or phrases to trigger responses from the Wechat chatbot. What we need to do is to create a set of standard keyword with answers structure on the platform without self-learning capacity.
Why Wechat chatbot matters
If your Chinese network matters your life and work, Wechat can automate communication and manage customer data. For example, when your Chinese network in Wechat asks you about Google or Japan Rakuten SEO fundamental knowledge, you could set up a series of standard tips and suggestions auto-reply, it’s a respectful way to show response in time. When a customer reserves a table in your restaurant through your Wechat official account, standard reservation flow, form, and response automates the process and make it done. When your prospect requests a demo to learn more about your business solution, Wechat chatbot automates to send them demo account login or white papers.
Also from the perspective of customer relationship management, Wechat is a powerful CRM platform where you can learn insight and optimize your product and marketing based on data. From millions of daily conversation data and interaction happens in Wechat, you could categorize who are prospects and customers; Who are loyal customers and who have canceled the membership; What content have most engagement rate and when customers engage with your content.
4 Ingredients to Create Wechat NL AI Chatbot
There’re 4 core ingredients or elements to build up a Wechat NL AI chatbot. Here I take Python for an example:
- Cloud server (Select Python server environment)
- Wechat official account
- Chatbot set up files (Python)
- Nature Language Artificial Intelligence Platform
First of all – Connect Wechat OA with Cloud server
When we log into the Wechat account back end, we go to the developer section and markdown the appID and app secret key. We would use these 2 codes adding on our cloud server, so our server can recognize which is our Wechat account server.
Of course, if you are testing the connection to see the performance, you can make use of WeChat sandbox. It’s like a UAT environment. The cloud server would respond to this ID, based on any post from Wechat users in the future.
Meanwhile, we need to add the cloud server address, token ID, and domain URL to the Wechat account. So Wechat can know which server to post and receive response information. Basically, the connection is completed here.
Secondly – Select a Nature Language AI platform
In China, there’re two well-known NL AI platforms to consider. One is Tencent chatbot cloud and Tuling China. Both of them have options for voice AI models, such as shopping, entertainment, sport, etc. You could select the model according to your Wechat business purpose. In terms of pricing, Tuline is slightly lower.
NL AI platform connects with the cloud server, and it can receive post information from our Wechat account, analyze the voice or text messages, and respond to users immediately. Generally, the model you select for your business is capable to answer common questions and just take time to learn some specific answers that fit in your business unique.
Lastly – Create and prepare chatbot developing set-up files uploaded to the server
There’re many open source frameworks and skeleton files in Github. You could refer chatterbot
In this file, you need to develop the code and create different user scenarios, when users would interact with your Wechat account. Basically, there’re 3 scenarios you must pay attention to
- New follower welcomed messages
- Real-time user-generated messages
- By default messages
The pros of NL AI must be its self-learning and optimization capacity. Thus, we just need to feed our business core information and Q&A logic, NL AI can help us reply to users’ messages.
However, the cons of NL AI is costly and time-consuming at the beginning. Although NL AI platforms provide different industry feature data-base that enable you to shorten the learning cycle, it needs time to run in as a whole with your business.
2 Ingredients to Create Wechat keyword-based response chatbot
There’re 2 core ingredients to create Wechat keyword-based response chatbot
- WeChat official account
- Skeleton and a full list of Keyword and replies
First of all – Wechat Built-in or External Server
Before we start to set up a keyword-based response chatbot, we need to find out which one to use
If we select to set up all set of keywords and corresponded answers in Wechat, we would reply on Wechat to finish the whole process and save the customer data in Wechat.
The pros of this approach is very easy and start fast. As the screen capture below, what we need to do is to upload and set up the keywords and reply messages in Wechat account, and it can publish right away.
For example, after I set up what’s Easy2Digital, users can receive an auto-reply of “About Easy2Digital“.
The cons of Wechat built-in is given functional model is limit, so if we need more customized functions, it might be a roadblock. Also, the account owner can’t filter out customer unique ID to connect with CRM and collect, manage, and utilize customer data.
On the other hand, if customer data and expandable applications are important to your business, I suggest you consider connecting an external server with Wechat. This approach is similar to NL AI response build-up. Below are 3 key ingredients
- Cloud server (Select a server environment, such as Php)
- Wechat official account
- Chatbot set up XML files
Some Wechat development platforms (For example WeiQing) provides a one-stop solution, that includes server hosting, Wechat account A/C, and chatbot development. It’s easy and convenient, but I would suggest thinking about the long term if you are going to manage your database in your own server. Before investing, please carefully check the terms and conditions of this section in the contract.
Secondly – Keyword and auto-replies
Different from NL AI self-learning capacity, the keyword-based response needs us to update the whole set of auto-reply time by time. So, below are my 3 directions to update
- All auto reply updated by manual
- Web search API feed + auto reply manual update
- All auto reply via API feed
Individual or SME might not have very complex product lines and business structure. So, I suggest you could update the auto reply in a fixed time. Here is a auto reply structure for your reference.
For some companies such as hotels, airline, etc, it has already built-in internal search engines, so I would suggest connecting Wechat server with search API directly. Thus, when your web server updates search engine capacity, Wechat auto-reply can be synced at the same time.
Wechat supports many different languages, but in my view, the main audience group is Chinese. So no matter which approaches you’re interested in to adopt, Simplified Chinese is the first option in Wechat chatbot.
If you like my content, please support us from below one of the ways, thanks for your support.
- Subscribe Easy2Digital Youtube channel.
- Follow Easy2Digital Facebook page
- Use the latest coupons and offer to purchase your favorite products if applicable (Saving your money and support me, win-win!)