FAQs Activity


The information presented here is an overview of Frequently Asked Questions (FAQs) activity. This document is especially useful if you are interested in understanding how artificial intelligence (AI) can be used to extend the capabilities of the FAQ activity. Also, external entries are added in some sections to provide technical details on how to integrate these capabilities in a Flow, which is mainly addressed to UXers, or whoever wants to build a Flow.

  • FAQs are conversational activities that allow end-users to ask a question and receive an automated answer as well as forward them to the appropriate flow if applicable.
  • FAQs is a set of AI-based capabilities and features integrated into the Yalo platform that allows workflows to understand natural language questions posed by users during the course of chatting with a conversational flow.
  • FAQs can be used as a directed activity (accessed via menu) or as a fallback when the user poses a question during the course of another activity.


A retailer would like their customers to ask questions about new services in a workflow on WhatsApp. From the menu, the user selects FAQs and then asks how to make a return. Either the answer doesn't exist in the FAQ, or the user does not understand the answer and asks another question.

Example: Flow Builder FAQs FlowExample: Flow Builder FAQs Flow

Example: Flow Builder FAQs Flow


  • Yalo Platform.
  • WhatsApp/Facebook User.

Dependencies and Technical Requirements for Semantic Search solution

  • Up to 40 questions and answers provided by the client via a CSV file. This will be your knowledge base.
  • Language support for Spanish, Portuguese, and English.
  • Questions that cannot be answered are redirected to the appropriate flow.
  • A .CSV file with 4 columns: ID, Question, Answer, Tags.
  • The “Tags” column should describe related topics or themes for the AI to recognize and make an association with the appropriate question and answer.
  • A customer could implement this themselves using a Flow Builder template.

Semantic Search

Semantic Search is the name of the text search engine integrated with the Yalo AI platform. It is called semantic search because it searches and retrieves text by comparing the semantic similarity of documents in a Knowledge Base and a user's question. We use state-of-the-art machine learning to compute the semantic similarity between documents and questions. This makes our approach much more powerful than commercial solutions such as Lucene or Elasticsearch.

Semantic Search is an unsupervised learning method. This means you don't need to train any models to start using it. The drawback is that it cannot have the same level of accuracy as a custom neural network. You can see Semantic Search as a generic method to search text. It has a good performance when it is tuned, but it will never be as good as a custom neural network

Custom Neural Network for FAQs

Another solution is the use of a Neural Network. Our AI platform can create custom state-of-the-art neural networks for FAQs given a training data set. The intended use case for custom neural networks is for those clients that want outstanding performance for FAQs and have the data needed to train a neural network.

The way it is configured is the same as intents: a CSV file with examples and category labels can be uploaded through Flow Builder.


Neural Networks for FAQs Are an Additional Service

Neural Networks are an add-on feature. Reach out to your Yalo representative if you would like to discuss adding a neural network to your service.

Did this page help you?