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 ManagerOli Morris, Principal Developer EvangelistJason Trent, Director of Product Management TOPIC Our first official Product Club of 2025 covered AI Agent Studio and how to leverage the practice of skill stacking to unlock tremendous potential value in your end-to-end processes. Here’s a rundown of the session: Jason refreshes us on Agentic Process Automation and AI Agent Studio. Jason deep dives some the latest features and updates to AI Agent Studio. Oli demonstrates how to stack AI Agents for maximum impact. Jason and Oli answer live audience questions. AI AGENT STUDIO: THE BASICS Agentic Process Automation (APA) is the third wave of AI capabilities that is able to take action and make decisions autonomously based on pre-defined goals. With Agentic Process Automation, you can automate up to 80% of entire business processes. Where humans needed to be involved before to create something or pull and analyze data, now AI Agents can fill that gap, allowing humans to move into a supervisory role. AI Agents are comprised of 2 critical parts: AI Skill + Action An AI Skill is something you build within Control Room that allows you to leverage a generative AI model grounded in your organizational data and fine tune the skill output with a prompt. An Action performs an automated task via UI or API automation. AI Agent Studio unlocks the ability to build enterprise AI Agents rapidly through: Democratized Development - All of our tooling is built natively into Control Room, so it's the same low-code, no-code development experience. You also won't need to jump into another set of data science or hyperscaler-specific tooling in order to connect to your automations. Enterprise knowledge (RAG) - This will improve the outcomes and accuracy of responses from models because it utilizes your data in a secure and safe way. AI Governance - Our deep level of governance enables you to track all the interactions from the models and provides dashboarding and tooling. NEW FEATURES TO AI AGENT STUDIO IN THE .35 RELEASE Custom Model Connections - We ship several pre-built connectors to different models on different hyperscalers, but because many customers are building their own models and new models are being introduced all the time, this feature allows you to connect to whatever you model you wantand bring it natively into AI Agent Studio. Then, you’ll benefit from role-based access and control, the ability to reuse those models across your automations, as well as AI governance tracking. AI Skill Templates - We’ve shipped 20+ pre-built AI Skill templates across service ops, finance, HR, marketing, etc. to give you a jumpstart with generative AI prompting. You can also turn any skills you build into a template for your team to use internally. Grounding LLMs with Automation Anywhere’s native Enterprise Knowledge RAG - Upload your content to improve the accuracy of your responses and reduce hallucinations. DEMO: STACKING VALUE WITH MULTIPLE AI AGENTS Let's run through a hypothetical support ticket. First, we want to think about how a typical customer support person might handle an email support request from start to finish. Open the email, read it, and classify it. Today Mr. Martinez has an issue with his Magic Mop 2000 leaving streaks, so the support person would classify this as a Product Support issue. This is important to determine so they know how to proceed with next steps. Next the customer support person extracts the data from the email in order to research the issue Mr. Martinez mentioned and determine the best resolution. They put that into the knowledge base to find the most relevant article to share with the customer to resolve their query. Once they locate a solution, they would put that information back into a ticket and respond to the customer. In this case Mr. Martinez should be sure to use a branded cleaner fluid with his mop. Now, if we think about everything that's just taken place, we can map that to capabilities that generative AI unlocks. Let's apply those to what a human is doing and gradually begin to stack them with AI Agents. Because once we start stacking AI Agents, that's when we can begin to stack value and achieve powerful outcomes. Triage Agent - Translates and classifies the support ticket. Maybe we've got customers emailing from all over the world in Japanese and in Spanish, so we could get an LLM to translate that. Or if there's a a heavy amount of technical jargon, the LLM and generative AI could translate that into simpler terms. This could help drive improved customer experience and allow the customer support person to spend less time having to work out query details, as well as eliminate the need for dedicated customer support people who specialize in certain languages. Research Agent - Extracts data, transforms data, and researches a solution. Maybe we ping a knowledge base which returns 5 results. The LLM could sort through those results and determine which is the most suitable to share with the customer. This would significantly speed up the resolution. Alternatively, if it's not the AI Agent recommending the best solution, the AI Agent could provide the customer support person with the top five knowledge base articles to review, saving the customer support person valuable time digging through hundreds of articles. Content Agent - Summarizes a solution for the customer to easily understand and drafts an email reply. This AI Agent could present a draft for the customer support person to send, as well as alternate draft options, so they could remain the human decision-maker or the human-in-the-loop, but ultimately the AI Agent would speed up the resolution and the job of crafting an email reply. Beyond this solution, perhaps over time you leverage generative AI to observe that Mr. Martinez has bought quite a few of these mops before and he's been really vocal about how great it's been on social media. So maybe instead of recommending the floor cleaner to him, we send it to him free of charge because he's such a great customer. You can start building in and stacking this level of knowledge to add tremendous value, which would likely be out of the question for your existing customer support set up. PATHFINDER ANNOUNCEMENTS We’re introducing the new role of Community Captain to our Pathfinder Community! As a Community Captain, you'll have the opportunity to lead discussions and build connections around topics that matter most to you, whether that's your industry expertise, professional role, or specific automation interests. If you're interested in becoming a Community Captain, click here to learn more. [NEED LINK] Be sure to subscribe to our Events page to receive updates when we post new events. Some events this year will be first come, first serve, so sign up to get those notifications and be sure you don’t miss out on anything you hope to attend. If you haven't already completed or updated your community profile information, please do so to become a verified member and receive a purple check mark next to your profile picture. This will help us customize your entire community experience! 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 and Oli, 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: In which version is this available--the AI skills templates? A: We just shipped this in our most recent release where we call it the .35 version internally, but it should have hit all of our cloud and on-prem customers in January. Q: Where can I read more about how to write AI training skills? A: No model is great at all tasks. There are definitely some models that are better at certain workloads or types of tasks that than others. With AI Agent Studio, we provide you the ability to evaluate and test different models with the same prompt. So even if you're not sure which model to use, you probably have access to at least one, two, or three different models in your organization, I would say to test all of them. As far as writing the actual prompt, there's lots of great resources. We're actually going to be bringing new prompting capabilities in our next release, which will come out in Spring, to make the accuracy of your prompts even better. We also have a course on AA University called Prompt Engineering for Developers, and we'll covering a lot of this in AI Training Camp as well. Q: Can we expect this to be tested from Community Edition in the near future? A: For AI Training Camp we’re testing now to make sure that will be available to users who choose to follow along in their Control Rooms, including the Community Edition. I can't comment yet as to whether you'll be able to do that from the Community Edition, but there's certainly a lot of stuff you can do in the community that would enable you to start playing around with Agentic Process Automation for sure. Q: Will it create a dynamic flow based on the user query, meaning the Agentic AI framework? A: Yes, that is something that we are building out in the training camp. You will learn about building prompts for generative AI and at the end, we'll be pulling those all together to do a fully end to end build, which covers the cases like the support tickets that I mentioned earlier. And we’ll be doing it two ways: one will be a deterministic way where you can stack AI skills together but it doesn’t put decision making in the hands of the LLMs, so the second will be an autonomous build where the LLM will have full responsibility to decide which tools or which agents it wants to call in order to answer the topic of the customer support. But yeah, you end up at a point where ultimately the prompts you have in your decision making agent is the orchestrator because that dictates what tools or what agents the decision making or orchestrating agent calls. Q: Follow-up question—does it have capability to take action based on instructions and goals? A: Absolutely yes. The structure of Agentic Process Automation is very much that you give it the context of what tools it has available to it, and if you've used Process Composer before, you'll know when an API task or a task bar runs, you can pass that state to the next action or even backup to a previous action. So you can give it the context of what tools it has available, but also give it the knowledge of what's previously happened. All of this can be achieved just through really solid prompt engineering framework. Q: How can I take part in AI Training Camp? A: Keep an eye out on the Pathfinder Community Events page. We don’t have a specific date yet, but it will be in the next couple of months, so check back there for an announcement and updates. Q: Is AI Agent Studio released and to whom is it released? A: Yes, it is generally available and we are on our 4th or so update. Reach out to your local account team to have them get you plugged into how to take advantage of it.