FAQ automation consists in improving your client's experience online by using FAQ chatbots to answer frequently asked questions.
Automation of FAQ (frequently asked questions) is the most popular use case for chatbots. Regardless of your industry or your scale, FAQ automation is a must and a great way to get started with chatbot and artificial intelligence. This should most definitely be on your radar.
It’s the fastest way to implement a conversational solution and create value for your employee and your customers. If you didn’t think of using FAQ chatbots before, this is the best place to start and experiment with a platform.
It’s easy to set up, doesn’t require any integration and allows you to test the natural language understanding (NLU) of the platform. Not all NLUs are created equal and it’s worth testing it before investing too much time.
To make sure you get the most out of a FAQ chatbot, it’s imperative that you build it the right way. I have a bunch of best practices for you to get started. It’s like building anything else in this world, foundation is everything. If you start off right, you are on track to change positively the way people interact with your business from the outside, but from inside as well.
The scope of your FAQ chatbot will be determined by your goals and your audience. These two elements should remain top of mind while you build your chatbot because they will impact the way you structure it and where you are going to deploy it.
First, identify your audience. Who’s this FAQ chatbot for? Your customers, your employees? This will help you figure out where you should deploy it and how you should write your utterances – we’ll come to that later.
It’s the first step of almost everything in communication, your understanding of the audience can, and will make or break the FAQ chatbot.
Now that you know your audience, you should know what they need from a FAQ chatbot to find it helpful. Focus on one single goal. Be careful here because it’s a common pitfall. Scoping too broad can make your FAQ chatbot difficult to create value.
It can be to help your employee with human resources-related questions or maybe you want to help them with something more specific like employee benefits.
What your goal can’t be?
Helping your employee by answering just about every question they may have about your organization.
Got scope? Moving on…
Topic is a concept we use to group FAQs and workflows together when building a FAQ chatbot. It helps you organize your content in the studio, but it also helps the NLU with the context of a user input to classify it. If two questions are similar, but have different topics the NLU will first classify the user input in a topic therefore allowing to choose the right answer.
How you structure your topics will depend on the scope of your FAQ chatbot. If you have a few questions per department, your topics could be named after your organization’s departments (HR, IT, Sales, etc.). If on the other end, you have a lot of questions for a single department, your topics could be named after the different functions of HR (Payroll, Vacation, Insurance, etc.).
Remember what your goal is and what is the content required to help your audience achieve that. What do your audience need to achieve the goal?
*French words for twenty-hundred.
The law of vingt-cent states that you should limit your FAQ chatbot to twenty topics and a grand total of one hundred FAQs. If it’s more than that you should build a delegation architecture to make sure every FAQ chatbot is performing to their best.
We came up with this after building hundreds of conversational assistants. It’s a rule of thumb to help structure your content in a way that’s maintainable and allows for NLU maximum efficiency.
Having 100 questions can easily be translated into having 1000 utterances if not more over the course of a few months. This is why we recommend building multiple conversational assistants and FAQ chatbots just like you would have multiple apps to manage multiple departments.
As I mentioned, this is a rule of thumb for Botpress NLU. It’s a good thing to keep in mind, but it may also not fit your FAQ chatbot very well. The numbers are arbitrary and going above the law wouldn’t mean jail time. Make sure to put some time aside to think about your structure.
In the context of conversational AI and FAQ chatbots, utterances are variants of a question or intention. If you want to know the current weather, you can ask in a dozen different ways.
Those are your utterances and as you can see they can come in different forms from a user.
They are important because they are used to train the natural language understanding which is essentially how the FAQ chatbot understands the user input. Having great utterances will result in a better FAQ chatbot and better conversational experience.
Here is a list of FAQ chatbot guidelines:
Note: Don’t worry if you don’t cover the full spectrum, the misunderstood module will help you catch those utterances and assign them with the correct Q&A. We all have this tendency of making things perfect before we release, but for a FAQ chatbot it is just counterproductive. The faster you get real user interaction, the faster you will ramp up its understanding skills.
If you'd like to get started, you can go ahead and download our latest version. Then, head to the documentation and follow the steps to get started. Now, it's time to build your chatbot using these guidelines.
It will help you dodge the major pitfalls of building a FAQ chatbot and create the most value for the end user. As I mentioned at the beginning of this post, FAQ automation is a must and it’s most definitely the easiest way to get started with chatbots.
If you think this post is valuable, please share it with your friends and colleagues.
Discover the capabilities of NLU software and the advances it has made to bridge the communicational gap between humans and machines.
Governments need chatbots for all the reasons that businesses do. They need chatbots...
RPA refers to software robots that run virtually and automate digital workplace tasks such as data entry.