Challenges with Deploying DialogFlow

by Jeow Li Huan

Share article

Huge team managing chatbots

Dialogflow is a natural language understanding platform used to design and integrate a conversational user interface. It has a very good offer, of 1-year free trial, for you to start building your chatbot. It is tempting to jump on it without much thought.

Upon starting, you’ll be greeted with incomprehensible terminologies such as intents, entities and contexts. DialogFlow is designed for a technical user in mind and would require prior training. Without this training, it can be difficult to get DialogFlow up and running, let alone create a functional conversational chatbot for users. You spend hours going through the tutorial just going through this first hurdle.

You create your first intent and you are faced with the next challenge. You may have an idea of what the interaction between the user and the chatbot should be like. However, DialogFlow is not designed to follow this natural flow. Instead, you need to create an intent for each chatbot response and use context to link them up. Often times, you may find yourself navigating in a blind, not knowing whether the intents have been correctly connected. This is further complicated by the “variable-first” design, which distracts you to write natural language processing rules for each intent even before creating the flow to present your content.

Updating DialogFlow to keep up with changing business needs is very hard without an overview of how the existing chatbot flow. Understanding DialogFlow analytics to get actionable improvements to make to the chatbot involves following every visitor flow to see what users searched is difficult. This difficulty causes a lot of chatbots not to be updated or expanded to cover more use cases.

In summary, it is better to think long term a choose a chatbot solution that is easy to learn and maintain, has a lower total cost of ownership and is able to grow with your changing business needs.

« Unlocking the Power of a Customized Digital Custo
mer Servicing ‘Hire’
Google Already Had a ChatGPT Rival a Year Earlier, but Cho
se Not to Release It »