Before embarking on the Week 4 Developer Track end-to-end build, we’ve included a video and some FAQs to ensure your Control Room is setup and you’re ready to go!

FAQs

Q. How do I import the tools into the Control Room?
A. Please watch the video above from 01:00.

Q. My Control Room looks different to how yours is in the video...
A. Please ensure you’ve completed weeks 1 to 3 before embarking on this build as we reuse some of those assets. Also, make sure you’ve created the autonomous and deterministic folders as per the video above. But, this shouldn’t affect how you build your final process if you follow the logic in the end-to-end video.

Q. I’ve run the autonomous build and I’ve hit an infinite loop in Automation Co-Pilot. How do I stop it and how did it occur?
A. This means that you’ve likely missed a step to write the tool state into the XML strToolState variable and the Agent is repeatedly trying to run the same tool. With the process running, click on “All requests” on the left:

Then on the running process, mouse over the ellipsis and click the red trash can icon:

Q. Some of my skills have the prompts in the system prompt field rather than the user prompt field. Does this matter?
A. No, this shouldn’t affect how your AI skills run, but please test them as I do in the end-to-end video.

Q. In the end to end build booklet, I’m unsure when to use the “Tool Execution States”. Can you explain?
A. Sure. At the start of the end-to-end video, I explain what we're going to be building in the autonomous build. And I talk about how we need to be able to store the state and the outputs of any executed tools into short-term memory. Remember that LLMs are stateless, so they can't remember what they've done. And we do this by using the strToolState variable. For the purposes of this video, I chose to encode this information into XML because it plays so nicely with Automation Anywhere and a lot of the foundation models which you may use in your APA processes, also understand it really well. So the tool execution states in the booklet just provide you with various examples of how the strToolState variable may look after certain tools have executed. So in summary, think of it as a snapshot into how the tool state may look at a given point in time.

Q. The “Create Return” (or “Create Sale”) tool didn’t work as expected. Why?
A. Firstly, your builds should still work in spite of this. I’ve built provisions into the tools to account for any errors occurring and the point of this end to end build is to focus on how Agents can work autonomously. Regarding the issue, this could be because your Browser Extension has not been installed properly and/or JavaScript hasn’t been enabled. Please ensure you’ve ran the ERIC process successfully to debug your connectivity to the Control Room.

Q. Do I need to complete the deterministic build too?
A. No. This was purely to show you an alternative way of achieving the same outcome i.e. automating the customer support process with Generative AI.

Q. I don’t have access to Automator AI for the deterministic build in Process Composer...
A. Don’t worry. This is not required. I was merely showing you how powerful it can be to use to build out your processes and automations.

o​@Oli.Morris will be updating the FAQs throughout next week]