In Virtual Agent Designer, use the Input Collector control to organize your topic flows and leverage generative AI for creating seamless user interactions. In our example, we've built a topic flow for gathering information from users to fulfill a T-shirt order. We need to know the user's preferred
T-shirt style, their size and what message they want silkscreened on the front of the shirt. In the Topic Properties tab, we selected LLM for the model type. This allows us to use generative AI in our topic. The first three nodes are LLM enabled user input fields. In the detailed description field, we tell the LLM what specific response we need from the user. By default, the user prompt
will be generated by the LLM. We could override this and enter a message to display, but using the LLM's ability to generate messages streamlines the process and frees up our time. We added a text node to generate a response to the user after they answered the questions. In the tell the LLM how to respond field, we entered a directive. Acknowledge that the user has made the following selections. Then we listed each input variable we collected and added the corresponding data pill from the data pill picker. Finally, we added a text response to confirm to the user that their selected options have been saved. Now let's test our topic flow. We see the user prompts that the LLM generated for us. Now let's see how adding an input collector node to gather all our T-shirt options streamlines our topic flow, adds a summary at the end, and creates
a seamless experience for our users. The input collector acts as a wrapper
for these other user input controls. We just drag and drop them onto the collector block. By default, you can add up to five input controls, but this can be changed by editing the VA input collector max nodes system property. The data and the input fields hasn't changed since
we added them to the input collector. In fact, we can drag them back out,
drag them back in, and reorder them as we like. We can also collapse the input collector with one click, which is handy when you have a long topic flow. The confirmation message is active by default and will display when all the inputs for this collector are received. With the input collectors confirmation message, we won't need the text response nodes
at the end of the flow, so we delete them. Now let's see what happens when we test the flow after adding the input collector. Users can change their choice for any of the inputs over the course of the entire conversation. Here we simulated a user changing their mind twice. The changes are seamlessly
blended into the conversation, giving the user a friendly and receptive
experience with virtual agent. And here's the confirmation message
allowing the user to verify their choices. As we seen, the input collector control
provides you with many benefits as opposed to stringing together multiple input nodes. As their building topic flows, topic authors can collapse them to
streamline the topic flow on the canvas. Generative AI can build user prompts, freeing up topic authors to focus on developing the topic flow and will generate a summary message when all the required inputs are collected. End users can confirm their choices
or make changes without having to redo the conversation, which allows for conversations to flow more naturally. Try using the input collector to help you better organize your topic flows and leverage the power of generative AI to streamline your Virtual Agent development process. For more information, see our Product documentation or knowledge base. Or ask a question in the ServiceNow Community.