Building your chatbots: existing solutions and opportunities

Chatbots are the new apps. As a new conversational user interface, many articles stated that there’s a huge opportunity for developers and businesses and urged that everyone should start to build up their chatbots. This is not what I want to discuss in this article, though. Here are some useful articles that tell you why it is important for you to build chatbots:

This article summarizes the difficulties and inconveniences for chatbot developers are facing today. As a chatbot developer, you can make your bots from scratch by utilizing these platforms’ API or SDK (which is the best way I recommend):


However, for some businesses that need to rapidly produce an MVP product to validate your idea via chatbots, using existing chatbot builders is obviously a better choice for you. Check out the following services that provide opportunities for those without any programming background to your own chatbots:


I’ve tried every one of them above when creating my own chatbots; however, I think there’s still room for improvement in terms of the overall user experiences for them as chatbot builders. Here are some things I believe chatbot builders could and should improve in near future.


Usually, a successful bot is done by a close collaboration of a group of product designers, developers, and marketers. It’s a pity that most of the chatbot builders today are designed ONLY for developers. It seems to me that the existing solutions still lack interfaces that are easy to use enough for different teams to work together. Here are some things I think an ideal workflow should be like: there’s a GUI for designers to easily drag and drop logic blocks or messages to finish a flowchart on canvas. Developers then can connect the webhook and implement required functionalities based on the flowchart. Then finally, an easy-to-use dashboard for marketers to monitor performance and collect feedbacks.

I’ve noticed that there are some companies working on these features, such as:

I look forward to using a powerful collaboration feature soon.


Connecting analytics to products is already a common sense among web and app developers for long, and sure the same applies to chatbots too. Chatbot analytics is more complicated than web and app analytics. Since the interaction between user and chatbot includes three major parts: user interface, logics, and conversation; page view and message count are not the only things you should worry about anymore.

For a service as complicated as chatbot, what traditional analytics tools are offering is way far from enough. New, specialized analytics that measures user engagement and conversational sentiments are needed.

To me, a good chatbot analytics tool should monitor all conversations between the bot and its users so the bot owner knows the entire customer journey, and when human agents should join conversations if needed. One can then take different actions towards each segmented customers and leverage on data to help increase user engagement. is a choice for you to use as chatbot analytics.


I am introducing two types of bot testing in this section.

1. Functionality test

Some builders are equipped with GUI-based live previewing. That is, you can use its graphic interface to enter messages as a user and see how the real-time bot-user interactions would be like when the bot’s published. An example of this kind is

The other type of solution is script-based testing for chatbots. This is a lot like what we’ve been doing when conducting product testings — you input commands to provided scripts and see what prints out on the screen. is a testing tool designed in this fashion.

2. A/B test

Just like in web and app service, A/B test is almost the mandatory step to take during one’s development cycle. A well-performed A/B test should be able to tell you which conversation style, media form, and even flow design is better for your business.

Unfortunately, there isn’t any available tool for chatbot A/B testing yet.

Hybrid bot support

A hybrid bot is powered by both real human agents and an AI-based bot. By adopting a hybrid bot, eCommerce companies can effectively decrease the number of human agents needed for customer service. Sounds like a fantasy.

The real challenge, however, is how to define the best timing to transfer the conversation from AI-based bot to a real human agent seamlessly. This is actually a problem we are tackling with here at By applying sentiment analysis, we automatically notify human agents when the bot-generated conversation starts to look confusing for customers (or when they seem pissed, of course.) Developers can also define their interruption timing to notify agents by inserting one line of codes from botimize’s API. This way, your agent can take over the situation immediately and save the day!


And more…

Now that you know where the opportunities are. I still have many interesting thoughts to share in the future article. For example, does anyone think about to have a GitHub for chatbots? While the number of platforms is increasing (Facebook messenger, Slack, Telegram… and more to come), it’s more and more important to have a place to reuse others’ frameworks, tools, and designs to accelerate the development process. Natural language processing is another key element in chatbot development. It’s worth writing another article on it.

We can see that there are many chatbot-related services coming up (e.g.,,, chatfuel, and more), but no one is monopolizing the market. So come and join the chatbot community and change the world now. It’s not too late.

About us provides analytics, A/B testing & optimization for your chatbot! You are welcome to send me an email at for any discussion.