My main use cases for Automation Anywhere include enterprise use cases in travel, banking, and supply chain. It's used for automating back-office business processes that simplify the work of day-to-day business operations.
The AI Agent Studio plays a crucial role in our automation processes as we use it to create autonomous agents, specifically back-office autonomous agents. We use it to automate processes where AI agents call the RPA bots for automation. The bots will call, and the agents will be calling the bots for the action part, while decision-making is done by AI agents. Wherever decision-making is involved, the agents take the lead. If there are pure actions to be performed, an AI agent will call an RPA bot. If there is any knowledge discovery or details to be searched, then it uses context routing or a RAG index. Finally, whatever an AI agent has done goes through a human in the loop for final verification by a human reviewer. This is a business process and AI agents are new, so we have introduced human verification for any actions done by AI agents. It goes through a human review before actually being performed in core enterprise systems such as SAP, ERP systems, booking systems, or insurance claim systems.
The main challenge or objective I was looking to solve with agentic process automation is related to tasks which cannot be purely rule-based automation, such as reading emails or going through content which is more unstructured in nature. When referring to unstructured nature, it could be contract documents, mortgage documents, reading mail messages, or chat messages. In all those areas, we cannot employ RPA bots to perform the automation because RPA bots tend to expect the data to be in a more predictable, deterministic, and more structured form or a semi-structured form.