Use natural language to create complex Processes
We're excited to release the newest version of Processes, which allow you to train your AI Agent to solve problems that require multiple steps and logic.
Now, instead of using blocks, you can write out multi-step instructions for your AI Agent to execute, complete with conditional logic and referencing Actions you've already configured. This way, you can use the same language you would use training a human agent to dictate what your AI Agent does.
This change also comes with some additional minor enhancements:
-
For Actions, we've added a checkbox that restricts your AI Agent to only running the Action when a Process references it. By default, the checkbox is selected, but you can clear it to allow your AI Agent to run the Action on its own.
-
We've added some new items to the reasoning log that indicate when your AI Agent ran a Process, or when it needed to gather information from the customer to run a Process.
-
We've added auto-capture variables to handoffs and other areas of the dashboard where you build content with blocks. That way, you can define information that your AI Agent needs to execute the block (e.g., the customer's accuont number), and your AI Agent can autonomously ask the customer for that information and save it in the variable.
For more information on our enhancements to Processes, see Create workflows with Processes.