Book a Demo

Introduction to Puzzel BOT trainer

Puzzel bot trainer delivers different levels of building a chatbot, from simple text answer FAQ-bots to advanced bots containing workflows based on data variables collected from the customer journeys, context variables and custom actions from REST APIs.

Puzzel Smart bot is a multi-purpose chatbot that can be used for a diverse range of customer needs. At Puzzel we aim to deliver a no-code, easy to build, low maintenance chatbot that focus on improving customer service experiences, lowering agent handling time and with a high degree of solution.

Puzzel Smart bot is an extension to Puzzel Digital Engagement and takes advantage of all the tools and features as a human agent has access to.

Example use cases:

FAQ-bot that aims to answer repetitive questions by learning the bot to understand human intents. This is done by building a strong NLU-model in the bot trainer tool.
First line-bot that aims to collect as much fundamental information as possible before handing over the customer to an agent. This is done by building qualifying workflows where the bot prompts the user with necessary questions.
Hybrid FAQ-Firstline bot combines the two first alternatives and is implemented using digital engagements rule engine. Enabling the bot to appear differently depending on the customer journey. Sometimes letting the customer ask questions, and sometimes asking the customer questions to provide information to the contact center.
Bots integrated towards REST-apis. Puzzel Smart bot can be integrated towards customers rest API’s. This is done inside of the bot trainer tool and is done to read and write data from third party systems.

NLU model

The Puzzel Smart bot NLU is a conversational AI that works with intents and entities. Where the intent is the fundamental parts of the bot. The Natural language understanding engine is what enables the bot to understand human queries.

The bot is trained by simply inserting a few training sentences in the training model.

By creating intents, you are building NLU-models in a matter of hours. The NLU-model can then be used to answer simple questions to performing advanced automated workflows to help guide and navigate the end user on a web site.

The NLU understands all possible languages, even made up languages if one would like. Puzzel Smart bot can handle an infinite number of intents.

Workflow Engine

The Smart bot workflow engine is what differs the Puzzel Smart bot from other conventional chatbots. The automation tools enable the user to build advanced workflows without any coding or programming background.

By using actions, a workflow is created like a flowchart that is executed top down.

A workflow can be triggered in different ways. It can be triggered by customer text queries. It can be triggered by buttons or by other linked workflows. This allows for creating dynamic customer experiences that lets the chatbot provide self service, increased customer satisfaction and a lower average handling time for agent.

A workflow can be a series of text answers, or it could be a series of actions that ask questions, gather data, provide information from an API and that take advantage of existing content on a webpage.

Features in Puzzel Smart Bot

Contextual intros

By using the Digital Engagement rule engine, our chatbot will engage the customer with different context depending on what customer journey the end customer is on.

Conversational AI

The Puzzel Smart bot includes a powerful NLU-engine that lets you build an intent model that is easy to maintain and that is easy on your in-house resources. The Bot trainer tool lets you analyze and train your bot in just a couple of minutes per day.

Automation tools

The Puzzel automation tools lets you automate part or whole customer journeys. By using the different actions that come out of the box, you can very easily build responses that lets the bot navigate the end user. That hi-lights information on you already existing webpages.

Use the built-in options action to create a guiding experience or collect data by asking questions in the conversation.

Easy to train

By analysing the customer queries you will improve the performance of the bot. It only takes a couple of minutes per day and is preferably done by a member of the contact centre staff.

API-ready

The Puzzel Bot Trainer tool has built in tools to integrate towards customer REST-API’s.

Multi-language support

The Puzzel Smart bot will easily identify you preferred language and switch seamlessly between language models.

Multi Intent understanding

The bot will identify more than one intent and provide options for the end user if two or more intents match the end user query.

Analyze

Analyze your bot performance in real time. Break out statistics in customer journeys, in recurring questions. Look at confidence levels and customer satisfaction score and more.

Getting Started

Logging in to BOT-trainer

The account to the bot trainer is created by Puzzel and they will provide the credentials to your account.

To login you open google chrome (recommended browser) and go to the following URL: https://botbuilder.puzzel.ai. You will be presented the following window.

If this is your first time logging into the platform. You will be prompted to change your password.

Logging out

To log out of your account you click your username in the top right corner. Then click logout.

The BOT-trainer Platform

This is the main/ home page of the bot trainer tool. Here you can see three different sections. The main menu shows you the options that you have available. This could differ depending on user permissions.

If the main menu is not visible it is probably hidden. Press the hamburger-menu in the top left corner.

If you are the administrator of more than one BOT you will be able to manage them from your account. You can change the bot instance in the top right corner by clicking on your name.

Main Menu

Training
Workflows	This is where you build your bots responses
Intents	This is where we build the Natural Language understanding (NLU) and train the bot
Links	Link-management
Test intents	Test more than one intent simultaneously to compare them against each other
Visual Trainer	Test what you’ve just trained the bot to do. This is a separate chat window to test out your bot and train it “live”
Statistics
Conversations	See live dialogues and train with real user input. Se statistics on conversations
Query log	Here you are able to view every text input made and analyze how well the bot respond
Intent statistics	Statistics on the bots accuracy
Workflow statistics	Statistics on the workflows that are used the most. These values may differ from intent statistics due to buttons and automated workflows
Administration
Settings	This is where you configure your bot

Back to Top