Can Copilot for PVA build a Card with Vertical Buttons?

Shows a gif how a topic gets generated by the Power Virtual Agent Copilot.

Now not to pad myself on the back too much, but I am somewhat of an expert when it comes to Vertical Buttons for Power Virtual Agents. Both for the classic designer using Bot Framework Composer and for the new authoring experience of Power Virtual Agents. And with the new prompt engineering feature Copilot coming to Power Virtual Agents I am wondering if my blogposts are now obsolete!

This post is of course about my first impressions using Copilot for Power Virtual Agents. It is also an excuse to quickly play around with it. I want to put this technology to a quick test as I did with Boost Conversation for Power Virtual Agent. I am very much an enthusiast of this technology and the productivity it can bring. So let’s put it to the test!

The First Prompt

Let’s now create our first ChatBot and ask it to write a starting topic. The starting topic will be the first question the Bot asks the user so our bot can direct the user to the correct topic such as:

  • Account question
  • Order information
  • Store information
  • Other

My first prompt to try this is as follows:
“After a user greets the bot, give him a hearty greeting in return and present him with 4 options to get started: Account question, Order information, Store information and Other.”

Shows the PVA Copilot starting point. The topic is "Greeting Copilot" and the prompt is: "After a user greets the bot, give him a hearty greeting in return and present him with 4 options to get started: Account question, Order information, Store information and Other."
First Prompt of PVA Copilot

And as you can see the result is exactly what we want. A nice welcoming question is generated. It even gets the tone we want to use.

Shows the Power Virtual Agent Canvas where the results are saved. One is with a message and the other presents the buttons the user can select
First Prompt Canvas Result
Shows the Power Virtual Agent test area where we test the first topic we build with Copilot
First Prompt Chat Result

Ask Copilot to change the options

One of the features of Copilot is that we can enhance selected nodes by instructing it with natural language. Let’s try to change the message and question to a card with Vertical Buttons with the following prompt.
“Can you change this to show the buttons vertically?”

Shows the Topic and the Copilot where 2 nodes are selected. The prompt reads: "Can you change this to show the buttons vertically?"
Second Prompt to update Nodes

Sadly this did not change any of the nodes. Copilot gives feedback that it did change something. This is slightly confusing as I know for a fact it did not change anything! I’m sure we will get a useful message in the future.

Shows that Copilot changed our Topic based on our prompt. However it did not change anything even though it said it did.
Result without feedback

The prompt I use is vague on purpose. I did not expect this to work as this prompt provides no context the model can relate to at all. Let’s try again with a more relevant prompt:
“Change these 2 nodes to a single card so and ask the question inside the card and add the choices as buttons”

Shows the PVA designer where a new message is added that posts an Adaptive Card.
Third Prompt Resulted in a Adaptive Card!
Shows the PVA test bot where our topic was changed by adding another card.
Third Result not entirely right

And even though it is not entirely right, I want to replace the nodes, Copilot did create an Adaptive Card for me. Which in itself is already really helpful.

The verdict

Okay I can and will not give a verdict on how good Copilot is. That would be mean by only playing out this scenario. I can give a verdict if Copilot can create Vertical Buttons. And I think the jury is still out on this one. If I’m playing the less knowledgeable bot builder which wants to change the choices vertically it will not help you. It does help you when you know what you need. My early verdict is that using Copilot will definitely be a large productivity boost.

Leave a Reply

Your email address will not be published. Required fields are marked *