Agentic Process Automation System (Now Certified for WorkSpaces)

Finance is where it starts in most companies. Other areas are slowly adopting it. In banking, automation is taking place in underwriting, loan process origination, loan documents, and servicing. Automation Anywhere  has established a strong legacy in finance and accounts. Given its extensive options, it can be implemented across multiple industries such as banking, capital markets, oil and gas industries, specifically in finance and accounts, and HR processes. It is particularly useful for repetitive and bulk document-based policies, setting up vendors, setting up employees, and numerous document-based aspects.

Automation Anywhere  has evolved significantly and upgraded itself to provide agentic AI and AI-based automation solutions for document automation. The product has matured considerably over time. We can create workflows that can call an API. We can include prompts in particular workflows for ChatGPT-related functions, connecting to an LLM and RAG to perform tasks. For document automation, modern features are available to train documents, ensuring high accuracy and repeatability over time.

The system is very easy to use. I recently completed a course in document automation, typically designed for people involved in coding and technical aspects. Though I understand coding comprehensively, I don't do actual coding. The course was very accessible. Currently, extensive coding isn't necessary due to the hybrid model incorporating GenAI aspects, low-code, no-code capabilities, APIs, and numerous pre-built objects in Automation Anywhere.

The features include GenAI-driven prompting methods and workflow creation capabilities. In these workflows, we can create decision boxes and call APIs without coding. We simply pull objects, drop them, connect them, and add minimal coding when needed. The most crucial aspect isn't coding but rather sizing the automation and fleshing out the details.

Automation Co-pilot takes notes and performs automated analysis. It can extract details from videos, summarize conversations, and provide detailed information. During calls, it identifies instructions and performs tasks such as preparing reports and reconciliation. Automation Anywhere can also connect with Microsoft Co-pilot. Through Co-pilot, real-time operations can be executed, allowing direct interaction between vendors and automation through this component.

The normal RPA  automations and integration with agentic AI are functioning effectively through Automation Anywhere. However, regarding document training, improvement is needed. Currently, it requires 10 or 12 formats for decision-making, which might not be optimal. The limited sample size affects automation efficacy, resulting in 50-60% accuracy. This needs refinement to achieve 75-80% accuracy with document understanding, reading, and processing various document types. The current accuracy rate of 60% for document-based automations requires human intervention. Focusing on improving this area would be beneficial.

I have been working in automation for the past six to seven years.

Maintenance is required for load handling, and some automations occasionally become idle due to sunsets. Customer support for error handling and troubleshooting is necessary. Currently, human dependency remains for these aspects.

The system is extensively used, with scalability depending on the number of licenses acquired. Each bot license handles a specific workload volume. Automation scale can be planned based on workload requirements, necessary downtime, and maintenance needs. Automation Anywhere accommodates these scaling requirements.

They are highly responsive to customer needs. I would rate their service as nine out of ten.

Positive

The setup process is quite straightforward. With agentic AI, we can incorporate APIs, integrate normal automated processes, and third-party tools. The workflow integration process is simplified, offering options to use existing objects or create new ones.

Implementation requirements depend on the specific process. Some processes require extensive in-depth study, while others involve simple automation. The complexity varies based on the process type and documents used.

Both time and cost benefits are achieved. While specific client details are confidential, the solution reduces time, increases accuracy, and decreases the number of process steps required to perform tasks. These improvements result in significant cost reductions.

From an RPA  perspective, it is affordable. Integration with LLM increases costs. Considerable focus is required on fine-tuning prompts, standardizing processes, and implementing effective guardrails. The expense of Automation Anywhere varies depending on implementation methods.

The tools include ERP  systems and existing automations. Automation Anywhere is platform agnostic and can work with multiple tools, including proprietary customer tools, in a collaborative manner.

Per bot, 0.10 FTE is achievable. Business analysis training requires approximately 20 hours.

Many on-premise features have migrated to cloud platforms to leverage agentic AI benefits. Automations and agents are predominantly deployed in public or private clouds, depending on organizational requirements.

Automation Anywhere continues evolving with a focus on business services. They need to conduct better research on document reading and improve the accuracy of agentic and GenAI automations to differentiate their product in the market.

I rate Automation Anywhere an 8 out of 10.