Workflows

To create a workflow, go into the main menu and under the category “Training”, go to Workflows.

On the left side of the window, you will then get the list of existing workflows. On the right are all the entities that you can use when building a workflow. This is your toolbox.

The most commonly used tools found on the right side of the screen when building workflows are:

Once you have created your workflow you will need to connect it with the corresponding intent. When that intent is triggered that means it will start the specific workflow.  
You go back to Intents in main menu. Select the intent you want to connect with your new workflow and click on the tab "RESPONSE"

When you go to RESPONSE, you will get the list of tools again. Here you select "Saved Workflow" and choose the workflow that your intent should respond with.
Now your workflow is connected with your intent and will be triggered when your intent is triggered.

To add your saved workflow, click add. Your changes are saved automatically.

Sometimes you must delete a part of your workflow. To do this you first select the node that you want to delete. When selecting a node its colour will become darker.

In the actions panel the selected block will open in the Edit tab. If you click on the settings drop down menu more options will appear.

Here you can update the position of the node in the workflow or perform deletion of the selected block by pressing the “Delete selected” button. You will then be prompted to confirm the deletion.

When adding a new node it will always be added at the very end of the workflow. If you want the added node to be moved to another position you can select the node and change the position by writing the new position under the Edit tab, the click on “Update position”. This also applies when changing the position of existing nodes.  

Another way to change position of your block is by clicking the  at the bottom of the workflow canvas. This will open a list of the block in the actions panel.

The bot-trainer will automatically create categories for you with which you can filter your workflows. You can filter by “and”, and “or” by clicking on the “&” or “|” icon in the filter categories field.

You can filter in different ways by using the and/or switch in the filter input field.

The “and/or” switch lets you filter on multiple filters. You can for example filter “Apply & Change”. This will include all workflows that contain Apply and Change. All others will be excluded.

If you filter with the or selector on the same words, you will include all workflows that contain apply or Change. This helps you look at more than one workflow category at the same time.

In this exercise we will create a basic workflow and connect it as a response to the demo-intent you created in the last exercise. The workflow will respond with a text-message and an option to restart the workflow.

1. Go to the main menu and select Training and then Workflows.
2. Click the sign to create a new workflow
3. Name your new workflow “Demo-Workflow”. Make sure that the checkbox is checked to log statistics for the workflow. This allows you the ability to export statistics from this workflow.  Click the “Add” button to continue. You now have a new workflow with its first node which states “Workflow Statistics”.
4. In the right toolbar menu, click “Conversation message” and input the following message without the quotation marks. “Hi! I’m a chatbot and this is a demo message!”
5. Click on “Add”.
6. Now in the right toolbar menu, click the Options icon. This will generate a button that will be presented right after the conversation message.
7. In the textbox labelled “Option: 1” enter the following. “Restart demo workflow”
8. Under the “Response action” menu, choose “Saved workflow”.
9. A new drop-down menu shows and you can input “Demo-Workflow”.
10. Now you have generated a button that starts your demo over when clicked.

Puzzel smart bot can take advantage of the use of entities. Entities are created by special markdown syntax.

Entities are used to avoid creating multiple intentions on the same thing. For example, a bank has several application processes. You can create an intent that lets the bot respond to all application-queries by defining entities such as "car loans", "mortgages", "private loans".

You do this in the training data by markdown. For example:

We are defining the entity within the brackets and applying the value to the variable withing the parentheses(apply). When using the variable, we are looking for the value of the variable “apply” and taking decisions depending on the value.

How to use entities in workflows

When you hit an intent with an associated entity, the bot saves the entity as a variable in its memory. To make logical decisions with the help of the entity, we need to build a workflow that works as a switch. We do this with the help of the action called "check if context is set".

This is an example of how the switch would look like using the loans entities example. We are looking for a value in each step, and if we hit a value in one of the Check for context nodes, we will branch out into the corresponding workflow.

We have covered how to build a simple NLU-model and some basic workflows that present responses in plain text. In this next section we will look at some more advanced but still fundamental tools in building a responsive workflows in the chatbot.

We will look at

• How to transfer a customer to a live agent.
• How to navigate and co-browse using the bot
• How to highlight specific elements on a webpage
• How to use options
• How to use Execute on time  
• Custom bot intros
• Enabling and disabling the chat input field

How to transfer a customer to a live agent

At some point your bot will have a hard time understanding. And at some point you don’t want the bot to answer a specific question but instead you want a live agent to take over. This is possible with the action called “Check group status and transfer to live agent”.

What this means is that the bot will look at the action and understand that it’s supposed to transfer the dialog to the service group that you have assigned the specific workflow.

Let’s say that a customer has a question about an invoice.

You have decided that questions about invoices are supposed to be handled by the live agents and your bot is to transfer the dialog.

Your live agents are part of different service groups and in this example let’s pretend that one of your service groups are called “economics”. What the bot will do is:

1. Identify the intent of an invoice.
2. Respond with a workflow that will say that the question is to be handled by a live agent.
3. Do a look up at the service group to see if the group is currently open and manned by agents.
4. If the group is open and manned by agents, the bot will transfer the dialog to the queue of the specific group.
5. If the group is busy (all agents unavailable), the bot will run a saved workflow.
6. If the group is closed it will display a save workflow.

When you create your transfer workflow it’s always good practise to add a basic “Conversation message” before you add the “Check group and transfer”-action. You don’t want your customers to be transferred without knowing what’s happening. Without the conversation message the bot will just transfer the customer to a group without the customer noticing it.

The above picture shows the properties for the check group and transfer action.

• Transfer to group: Choose which service group the bot will transfer to
• Workflow to Execute on Busy: Choose a saved workflow that will trigger when the chosen group is busy
• Workflow to Execute on Closed: Choose a saved workflow that will trigger when the chosen group is closed

How to navigate and co-browse

A good practise when building your bot is to help your customer to find information instead of just presenting the answer in plain text. To do this we can use the built in functionality called “Co-browsing”.

In the bot-trainer we call this action “website navigate” and basically what this does is to navigate the customer to an assigned URL.

To add this action to your workflow, you simple choose “website navigate” in the actions panel and type in the URL you want to navigate your customer to. And then click “ADD”.

This will add a new node to your workflow with your “website navigate”-action. So in this case the bot will state in a conversation message that it is a test message and that the customer will be navigated. And in the next step the bot will navigate the customer to the specified URL.

How to highlight an element on a webpage

Let’s say that you want to not only navigate your customer to the right page at your web. You also want to really point out the information that is relevant to the question. This could be, for example pointing at an application form or pointing at a product.

This is possible with the action called “Highlight element”. Highlight element will highlight an element of your choice with a yellow box. And in the same time scroll the customers browser to the correct stage of the page.

To use the “highlight element”-action you have to know how to copy the “x-path” of the specific element. The x-path is the unique identifier of a specific element on a webpage or in a web application. To find this unique ID, you have to open up the browser inspector.

From Chrome :

1. Right click "inspect" on the item you are trying to find the xpath
2. Right click on the highlighted area on the console.
3. Go to Copy xpath
   

How to use options

Options are used to navigate the chatbot with buttons without writing text to it. These are essentially buttons the customer can press and be redirected.

For example, you can present buttons as a menu when the chat bot starts. It is even recommended as this helps the customer to the right questions directly. Best practise is to present buttons with clear meaning as bot intro.

To add options to your workflows you go to the workflow builder view and add options from your toolbox.

1. Enter ‘’Option text’’. This is the text that will be stated on the button.
2. Chose response action. It’s recommended to use a saved workflow.
3. Use the  button to add more options
4. Use the   to modify the order of the options.
5. Click ADD to add the options to the workflow. This will create a new node.

Use the icon to enter more preferences.

How to use Execute on time

Execute on time is a feature that allows the bot to preform different depending on time of day or even dates.

A common example is that the bot greeting is different if it is used in the morning, during the day or in the evening. The bot can then greet the customer with for example “Good morning”, “Hello!” or “Good evening”.

Execute on date

You can add custom bot intros that executes depending on which URL your customer is on when initiating a chat session.

Let’s say that the customer is navigating to the customer service page and from there initiates a chat session. The bot will greet the customer with a custom message for the customer service and a custom menu of option buttons. If the customer is navigating the e-store the bot will prompt the customer a custom intro with option buttons on how to buy your articles.

To create a custom intro for your chatbot you need to create custom ‘’Opportunities’’, custom ‘’case types’’ and custom ‘’Solutions’’ in the platform. This is something that is done firstly by the administrators at Puzzel in collaboration with the customer.

Chat input field action

Let’s say that you want to create a workflow where you want to force the user to click through different options. Or you want to force the user to fill in an input field inside an inline HTML-template in the chat. Then it’s a good idea to disable the chat input field.

Remember to enable the input field after you have finished your workflow or task. It’s a bad practise to not let the user write queries when supposed to. The workflow below is what you see to the left. When form is submitted, the workflow will continue and enable the input field.

Setting intent model

The Puzzel Smartbot can handle more than one intent model. This is preferred if you are building a bot with more than one language. It’s then a good idea to separate the training data into two different intent models.

When starting the bot, you then use the action “Set intent model” to select which intent model to query in the dialogue. The action Set intent model can be used where ever one may want in any workflow.

The action is a node just like any other and therefore it suits best in the intro-workflows since we want to set the intent model as soon as the bot starts.