The Agentic solution is our AI solution, where Document Understanding uses an AI ML model. For Agentic Automation, we integrate with Large Language Models to achieve end-to-end automation.

I don't yet use Agentic AI. Part of the reason why we're attending a conference is to learn more about it.

"We are exploring a use case for agentic AI right now."

I assess Agentic AI's ability to extract data points as unclear because we use generative AI technically, not Agentic AI to extract data points.

We're starting to get into the Agentic processes, so that's been really cool seeing that, getting to know how it works and building out solutions with that.

The use cases for Agent Builder involve an automation that we're currently working on for our owner relations department. They receive a ton of emails asking about various information about their ownership, or maybe needing to pass off their interest to a family member, or they've sold their interest. So this agent will receive those emails, get the context of that email, get the sentiment, and be able to route that email to the correct person or department that needs to handle that request. So it could be a revenue question, it could be a land question, it could be all sorts of different questions.

My overall experience with using UiPath Agentic automation capabilities is that it was very easy and rapid to develop, and we achieved a more sophisticated end result than when we tried to use other AI tools to do a similar thing, which were not nearly as sophisticated or accurate.

We created what we call an email conversation agent that engages with a customer, vendor, or someone in an email conversation with a set goal of extracting specific data points for an action to be performed subsequently. I would assess Agentic AI's ability to extract data points as effective.

I currently use agentic AI. I assess agentic AI's ability to extract data points as effective, especially with all the capabilities and the new models that came in with the Anthropic model, which I appreciate greatly. It's developing rapidly with many new features, so I don't have to look for other platforms.

An example of time saved with Agentic AI would be taking customer information from numerous emails that would require manual human processing. Running an Agentic process takes minutes compared to what could take an adjuster anywhere from an hour to days depending on the information volume.

I use Agentic AI, and our current experience with its ability to extract data points has been doing really well. It requires fine-tuning, but the data we are getting back is within a threshold of between 85 and 90% accuracy.

Using the Agentic AI, I can process the documents easily and can just get the narrative that is present in the document to determine if we need to give the refund back or not.

From what I've seen while assessing Agentic Automation's ability to extract data points, it extracts the information really effectively, especially with Document Understanding and IAXP Document Understanding, which has been amazing with the way that we're able to use that and correctly identify the type of document that we're looking at at the time.

Do you use Agentic AI? If yes, how would you assess Agentic AI’s ability to extract data points? Please explain why. We don't yet use agentic.

As we are moving towards the Agentic, we have used Agent Builder and Maestro to automate one of our processes which was related to seat assignment. The agent was able to identify which seats are available based on the user preference and seat map, and it was able to assign the best-suited seat to the user.

We pushed Agentic AI to production in early September and so far, we've saved roughly 72 hours. It's processed 400 emails, specifically 400 purchase order emails that we use it for.

I am also exploring Agentic AI.

Agentic Automation in UiPath Platform has not been used yet, but it is something to look forward to, and a new certification is being pursued. There will likely be some use cases in the team to replicate that.

The use of agentic automation in UiPath Platform has transformed from rule-based logic using if statements to AI agents that analyze data to make decisions. This is particularly useful when dealing with unstructured data, such as invoices with different layouts. AI can classify and extract data accurately.

I am not using Agentic automation in any of my production environments, but I am using it for development purposes. I am exploring the platform on the Agentic side.

There is considerable work needed in Agentic AI where the LLM approach is taken from a third party. It is learning, but you need to reassess, reprogram, and redo the work. The learning process needs improvement so the agent can auto-learn. My main problem is that when I put an agent with a robot and it escalates to a human, I need the agent to assess and understand the problem-solving capabilities and the importance of solvability to determine if it can escalate a second time or not. We are not there yet with this functionality.

Regarding agentic automation, one of the challenges I sought to overcome by leveraging people, robots, and AI agents in automating processes was the difficulty of integrating multiple applications through UiPath Platform Agentic AI. Recently, UiPath Platform introduced UiPath Maestro, which has significantly resolved many of these problems. I have developed several Agentic AI processes using UiPath Maestro, and everything has worked well. Furthermore, with my prior experience in prompt engineering, UiPath Platform Agentic AI training has instructed me on writing prompts effectively, so I do not face any issues during the development of Agentic processes.

I have not started using the agentic automation in UiPath Platform because I took a course last year at a conference, and I think we have not got any use case for that.

I do not know how UiPath Platform can be improved. I believe UiPath Platform is the number one in its field in Agentic AI and RPA. However, in Agentic AI, even if it is the number one, there are some limitations. I believe in the future it will be much better.

I did not have a chance to try agentic automation on a real use case; I just started it by myself to get familiar with the technology, not diving deep into that, but I appreciate the idea of Maestro and how it looks. I really appreciate the approach of showing just a simple BPMN or diagram where you are solving your puzzle and collecting all the components in one place. It reminds me of my first RPA tool, Blue Prism, which I also really appreciated. I never faced any issues because I did not use it for a real use case, but it looks promising, especially how easily you can automate and create agents and context-based agents with RAG. I appreciate it and I am hoping that there will be opportunities in the nearest future where I can use something from Maestro or agentic automation or all together, which would be interesting.