Chatbots
In this guide you'll learn everything about how the technology behind your chatbots (YolaAI) works, the steps to setup, test and deploy your chatbot, and the most important tips for highest accuracy.
What is YolaAI?
YolaAI combines a Large Language Model (LLM) for generating dynamic text responses and intents for handling structured, predefined workflows. Together, they create a flexible and scalable solution for customer support automation and can reach a 99% accuracy when responding to your customers.
Large Language Model (LLM): Generates conversational text responses based on training sources like URLs, documents, or snippets. You can choosed your preferred LLM, such as GPT or Claude.
Intents: Handle structured workflows for specific customer requests that go beyond simple text responses. Examples include collecting information, validating inputs with APIs, or triggering actions like creating tickets or transferring to agents.
If you have both LLM and intents enabled, YolaAI will first look for a matching intent and if none matrches the question from the customer it will fallback to LLM and use the training sources to craft a response.
Steps to Train and Deploy a Chatbot
Follow these steps to create, train, and optimise your chatbot in Yola. Follow the link in each step for detailed guidance.
Use the default chatbot provided when you create an account, or set up a new one for specific use cases.
Add and configure the languages your chatbot will support.
Enable the LLM toggle to use dynamic, conversational responses powered by your preferred LLM, such as GPT or Claude. Tip: If you disable the LLM, your chatbot will rely on structured workflows and predefined responses and not on your training sources. Enable LLM for a more conversational experience.
Add content from URLs, documents, or text snippets to train your chatbot.
Tip: Ensure all critical information is in plain text since Yola cannot read text embedded in images.
Write clear, specific prompts to guide your chatbot’s tone, behaviour, and accuracy.
Tip: Test and refine prompts frequently to ensure optimal results.
Use intents for customer requests that require actions like:
Creating tickets
Agent takeover
Segmentation of VIPs on chat started to bypass the chatbot and speak to agents directly
Tip: Use chatbot actions like JSON API, Modify Variable, or Branch to build workflows tailored to your business needs.
Continuously test your chatbot after completing each step to ensure it responds and behaves as expected.
It's time to deploy your chatbot! If you're deploying to your website, you have endless options to customise the widget according to your branding and you can deploy the same chatbot to all apps we have available in a few simple steps.
Last updated