If you haven't heard about Product Club — the Pathfinder Community Product Club is a monthly virtual meetup led by Automation Anywhere product leaders that focuses on our latest proprietary product innovations. It offers a place for community members to stay informed, connect with product leaders, and gain insights into real-world applications of the latest innovations in intelligent automation.

P.S. If you can’t attend a meeting, no worries — we'll be dropping a recap of each month's session right here in our Product Club hub.

HOSTS

Lu Hunnicutt, Community Manager
Jason Trent, Director of Product Management
Rinku Sarkar, Director of Product Management
Nirmal Mathimaran, Lead Technical Writer - Dev Services
Max Cassidy, Senior Manager - Community & Pathfinder
 

TOPIC

December’s Product Club covered Agentic Process Automation and all the recent enhancements we’ve introduced into AI Agent Studio to help our customers automate with AI Agents with speed, ease, and security assurance!

Here’s a rundown of the session:

1. Jason provides a refresher on what makes an AI Agent.
2. Rinku dives into the governance features and tools in AI Agent Studio.
3. Rinku takes us through a demo of the new governance features.
4. Jason demos how to ground your models with RAG and how to access the new AI Skills templates.
5. Nirmal takes us through additional AI Agent Studio documentation on the docs portal.
6. Jason, Rinku, and Max answer live audience questions.
7. Questions about setting up Enterprise Knowledge were asked, and to assist, Max has created a helpful video explaining the process LINK.

THE FUTURE OF AUTOMATION IS AGENTIC!

Agentic Process Automation (APA) is the third wave of AI capabilities that is able to take action and make decisions autonomously based on predefined goals.

To refresh your memory, AI Agents are comprised of 2 critical parts: AI Skill + Action

1. An AI Skill is something you build within Control Room that allows you to leverage a generative AI model grounded in your organizational data—aka Retrieval Augmented Generation (RAG)—and fine tune the skill output with a prompt.
2. An Action performs an automated task via UI or API automation.

APA IMPROVEMENTS IN RECENT RELEASES

• RAG - The ability to ground generic LLMs with your proprietary enterprise data so AI is pulling data from your sources when you make requests to the model, significantly reducing hallucinations.
• Faster AI Agents - You can now embed AI Agents directly within API tasks, allowing agents to execute faster than ever.
• Governance - You can enforce the use of only approved models through our Model Connection management, as well as audit and monitor interactions and prompts with the models in real time. With our recent enhancements, you will have complete visibility into the risk posture as you implement the responsible AI framework within your organization.

GOVERNANCE DEMO

AI Governance Dashboard Features:

• On the AI Governance tab of your Automation Anywhere Home dashboard, at the top, you have an interactive visualization of the top models being utilized across your automation estate. By hovering over each bar, it will display a breakdown of how each model is being used.
• Directly below this is your Model Utilization table with a 360º view of your usage trends—what automations are using each model, what are the number of model calls you’re making, what devices are running, etc.—and all of this information is exportable into CSV for whatever offline purposes you may need.
• Below this table is a Token Consumption widget to help you visualize consumption data and optimize your costs as you’re using generative AI across your automations. 
• Then the table below this widget provides even more in-depth details into your token consumption. You can click on automations to drill down into a log view and click on a session to view model interaction details, prompts, responses, parameters, etc.

Code Analysis Enhancements:

• You can set AI governance rules, such as defining the vetted and approved models for use, that will flag as violations when you run a code analysis on any given automation. This puts you in complete control of what models developers can use within automations.

RAG DEMO

To recap the existing Model Connections feature—This is where you securely connect to generative AI models, then share that model connection in your organization for use across AI skills, AI agents, within your automations, and within your solutions. We support a number of leading models across all the major hyper scalers, Open AI, Azure, Google Vertex, Amazon Bedrock.

What’s new is that we’ve brought some new RAG capabilities into the models as well. So now, when you set up a model connection, you’ll notice a new option to “Choose a type.” Here we’ve introduced the option for a standard model, fine tuned model that’s been enriched with your own data, or a model grounded by knowledge base where you will see a number of options are readily available to select.

Within the next few weeks, we’ll roll out the capability to integrate with our AAI Enterprise Knowledge where you can upload upload content and use it as a RAG source. To set it up as a model connection type, simply select Automation Anywhere as the vendor and you’ll see Grounded by knowledge base automatically selected as the type since it’s the only model connection option we offer.

Another new enhancement makes it possible for you, as a customer, to register your own model definitions within Automation Anywhere and build against those. This comes because lot of you have told us you want to take advantage of models across the entire internet as soon as they're available—Hugging Face, for example. We've now introduced a new API that allows you to go and register your own types of models. To do this, use a tool called Postman to run an API call for, let's say Meta Llama on Hugging Face, and when you return to create a new model connection, you'll find Hugging Face as a vendor option and Meta Llama as a model option. As new models are coming out, we'll continue to make these Postman payloads available so you can start using those within your system as soon as possible.
 

AI SKILLS FROM TEMPLATE

We already have the capability to create automations from templates, starting with task bots and then we added process templates. Now, we've added the capability to create an AI Skill from templates. 

In your Repository, simply click the “Create” drop down at the top, select “From template” and then you’ll see the new AI Skill tab populated with 4 templates and we will be rolling out many, many more in the coming months. To explore what each skill does, click on the binoculars and a new tab with open for you to preview what the skill looks like.
 

THE 2025 AI AGENT STUDIO ROADMAP

1. Improve the AI Agent landing page experience in Control Room.
2. Introduce new system prompts to improve accuracy of AI Skills.
3. Invest heavily in prompt testing suite.
4. Enforce data masking rules for sensitive data (PII, PHI, PCI, etc.) within prompts and analyze any toxic content that might be embedded when using Generative AI models.

ADDITIONAL DOCUMENTATION

Automation Anywhere’s documentation portal is your go-to resource for comprehensive information about AI Agent Studio, and is constantly being updated with all recent enhancements, including incredibly helpful visuals. Everything Jason discussed today is nested under AI Tools (grounding models with RAG, creating custom model definitions, and creating AI Skills from a template), and everything Rinku discussed about governance is nested under AI Governance.

SESSION Q&A

Thank you to our audience for submitting their questions! Unfortunately, we aren’t always able to answer them all during the live session. We want to express our gratitude to our special co-hosts, Jason, Rinku, Nirmal, and Max, for providing their expertise and responses.

**Please note that all answers were shared live during the December 2024 meeting, and are subject to change. We strongly encourage you to contact your account management team for any specific licensing and pricing inquiries.

Q: When we upload documents to the knowledge base, is there a way to configure RAG architecture?
A: Yes, you can in many different instances. Specifically, Amazon Bedrock Knowledge Base and Google Cloud Vertex AI offer a lot of customization options whenever you want to go and configure both those RAG offerings. Automation Anywhere's Enterprise knowledge is a little bit more sort of focused on getting documents uploaded quickly. There's some small modifications and changes you can make, but by and large we actually take care of all a lot of that heavy lifting for you. And so really as a customer, you have almost infinite, you know, options on which of those different sort of approaches you want to take.

Q: Can we get some use cases to use Agentic Automation and any detailed courses to use Enterprise Knowledge product?
A: We have finished creating the learning video and are in the process of getting this available on AAU. We will post within the community once it is done. What I will do now is get it available on Youtube and will share the link in the chat.

Q: Can we get documents used in the tutorial as well as in the YouTube description to try hands on?
A: Yes, the use case that we've got, we have sample documentation that can be available and is all open source. So to really be able to test this and get hands on with the enterprise knowledge, when you watch the video that we put out there, you'll be able to build this a test case on open source documentation that you can easily find online. And you'll be able to see how you can really tailor your automations based on the data sets, just like Jason was showing in the demo today.

Q: Can we get the insights via API Call?
A: We currently don’t have direct API calls to gather the insights, however, the governance dashboards give you access to everything. You are also able to capture all runtime data during execution and use it as you wish.