Agentic Process Automation System (Now Certified for WorkSpaces)

Currently, our use case for the solution is majorly in IDP  solutions, including invoice processing, document processing, and exploring agentic solutions. Major issues that we are having involve multiple types of POs that come along and how to process them, how to get good structured data from them, and how to remove the hallucinations of AI models. These are the major scenarios, and after that, the processing in SAP and other platforms with that particular data follows, which is our automation and IDP  use case.

Document Automation plays a very good role for document processing. We were initially using IQ Bot, which is a product of Automation Anywhere , and we have now migrated to Document Automation. It helps us because in the previous version, we used to handle a lot of groups, and handling them was very difficult when we had multiple customers and a lot of templates of POs or invoices. Now we have one model, which is a smart solution that learns from what we give it and from the prompts and rules we use. It gives us better results, and we do not have to maintain it. It is also easy in live production because previously we needed to train in dev, then go to UAT and production. Now we can give validation two or three times in production and do not need to code it again and again.

We are using Document Automation on a large scale. For processing a manual PO and extracting its tags, it would take at least six to seven minutes for a person to do it manually. With Automation Anywhere , it is a very quick task because it extracts all the data in one go. If we send 30 POs, it will extract all of them at once, and then it will proceed with processing. One person can do all the work, and it saves a lot of time because it completes the task in a minute or two.

Automation Anywhere comprises multiple features, and it is a place where we can integrate a lot of solutions, which involve things such as automation solutions in Workday  or Salesforce , agentic automation, human in the loop, and other varieties of things. The feature I love most is how we can interact with a human in the loop within IDP solutions so that a particular person can validate that the extraction is correct and that the data going inside SAP or any finance system is accurate, and how they can interact with it. The processing reduces their work as well, which is also a good point.

The API has helped me achieve my automation goals by allowing us to do a lot of POCs so that we can interact with the whole process structure with human-in-the-loops because in our industry, data which is going in and coming out should be validated. This is helping by reducing a 10-person workload to one person who can validate and approve the work.

It is creating a lot of value for the organization because there is a very big team which used to do a lot of processes. For example, the research and development team used to find chips that have a lot of applications and determine what those applications are so they can perform R&D or innovation on chipset manufacturing. We created LLM-created solutions, and it is extracting what they used to do R&D with a lot of data manually, which took a lot of time. This automation has reduced the time, and whatever they used to do in a week is now completed within a day. This is an example of how we are reducing complex times with the automation.

Sometimes the challenges for me and my organization involve the infrastructure because agentic AI also needs infrastructure to be set up, and the current infrastructure setup is not very great to adopt it. Sometimes we have issues which take time to understand, and that is one of the challenges. Another challenge is that we are using it for IDP solutions, and AI and LLMs are still hallucinating and giving different results for the same thing. These are the things that we are still struggling with. Automation Anywhere support provides assistance whenever there is an issue. When we raise it with the team, they are very responsive and trying to resolve it by coming on a call, and they resolve it as soon as possible.

We are currently using the Autopilot capability of AI Automator AI , but not in a very vast way or in a major way because our automation is more of a scheduled version, so we are not using it much right now.

In the initial session, there were a lot of things that are in preview, such as A-Sera and A-Code, and there are some process discovery tools as well which are coming, which are actually very great. They are in preview, so I am waiting for them to come. From my previous experience, sometimes when a new product came, it took time to get matured. I would be expecting that when it launches, it comes as matured.

I have been using Automation Anywhere for the last seven years. Parallelly, I have done solutions with Power Automate , but as compared to this, I feel that I am more comfortable with Automation Anywhere.

We used to have a hypercare period when doing a deployment, and I see a gradual decline in the errors. Right now, if I talk about four years back, we used to monitor deployments for a month. Now we monitor for a week because we have already tested a lot. Generally, issues are not coming from Automation Anywhere; more of the issues are coming from the infrastructure, which is fine and usually happens.

Automation Anywhere provides support whenever there is an issue. When we raise it with the team, they are very responsive and try to resolve it by coming on a call, and they resolve it as soon as possible.

For small automations, I would prefer Power Automate  and similar solutions. For a complex one, I would go with Automation Anywhere.

It depends on the process. We are in the R&D space, and as I mentioned, the one-week work is done within a day, so it saved five days. For order processing, I would say that it is not very much time-saving because it is also taking time to extract and process, but it is giving us freedom in that the people who are doing the repetitive task are not doing that repetitive task now, and they are doing something else, maybe validation or something more productive.

For small automations, I would prefer Power Automate and similar solutions. For a complex one, I would go with Automation Anywhere.

The importance of AI governance in my organization is always crucial because of auditing, and we do not want any data to come out from there. Automation Anywhere gave us that particular security, and they are handling the data within their enterprise and not going outside. That is the first point which is why the organization is using Automation Anywhere.

We are using AI agentics and AI Agent Studio  for some of the processes. I would say that it is majorly in a POC stage, but as I mentioned, we are using it for some of our systems and the API with the agent connection as well.

The visibility is more because in any complex solutions, we used to take the help of it. Visibility for us is significant and it does create an impact because whenever we are stuck for some time and then use it, it is clearly helping us a lot for the creation of solutions.

The API has helped me achieve my automation goals by allowing us to do a lot of POCs so that we can interact with the whole process structure with human-in-the-loops because in our industry, data which is going in and coming out should be validated. This is helping by reducing a 10-person workload to one person who can validate and approve the work.

I do utilize Automation Anywhere Center of Excellence, CoE somewhat for solutions creation and if there is any complex way which we are not able to identify how we can resolve these things. We use that for support.

I give this review a rating of 8 out of 10.

Our main use cases involve automating finance-related bots, specifically finance invoices, intercompany receivables, and manual entries. All those types of bots we are automating.

The main automation we have built is for the finance team, which was receiving around 500 to 600 requests for manual general entries. For those entries, there was a particular request they used to get from the user, and then they had to validate that request before posting it in SAP and attaching the document in SAP, followed by communicating details of that request with the stakeholders. This was a time-consuming and repetitive task, so we built an end-to-end automation bot that monitors Outlook emails, monitors the shared mailbox, takes and validates the request, the user, and the attachment. Once that is done, it posts the document in SAP and sends a successful mail to the user. If there is any failure, the bot also sends the particular error message to the user to eliminate any communication gap between the user and the bot. This basically eliminates all the manual efforts of the users and increases the efficiency of the task process.

We are about to use Agentic Process Automation ; we are currently using it for a particular region and expanding that bot to other regions as well, alongside using it for Agentic Automation .

My experience with Automation Anywhere  is quite good because we can automate the whole end-to-end process with no human intervention. The main focus is that we are able to focus on higher-value activities rather than working on the same repetitive tasks, which is quite important and gives the highest impact to the business as well.

The main automation we have built for the finance team eliminates all the manual efforts of the users and increases the efficiency of the task process. We are able to focus on higher-value activities.

Currently, while working on the SAP part, I encounter some challenges, particularly not being able to extract the tree structure. There are some tree structures we are unable to extract, so we are finding difficulties with those objects in SAP. I think the SAP part can be improved.

I have been using Automation Anywhere  for around 3.8 years to build end-to-end automation RPA  bots.

Sometimes Automation Anywhere recorders might fail, and there could be package issues. I have faced those issues, but not major ones—just minor ones that I was able to resolve myself. I received support from Automation Anywhere too, and we used to raise tickets, which helped a lot, with hours spent solving the issues over a call.

In terms of Document Automation, I am not using it in the current organization, but I have used it previously for the invoicing process. It used to capture documents in PDF format, extracting data from tables, including quantities of the invoices and the final amounts, and sent the mail to the business user. For Document Automation, we used to get around 15 to 20 requests monthly for documents, saving about 10 minutes per document, which amounts to a 40% efficiency saving.

I would rate this solution with a score of 8 out of 10.