Advanced Automation: Intelligent Automation and Agentic AI Concepts

Why is this important?

Whereas traditional RPA offers significant efficiency gains by automating structured, rules-based tasks, it is increasingly important for students to understand how automation frameworks are evolving. Intelligent Automation (IA) and Agentic AI represent the next frontier, integrating machine learning, document intelligence, and generative AI into workflows. This class introduces students to the expanding capabilities of automation systems, motivating a transition from purely procedural bots to adaptive, decision-capable agents capable of completing complex sequences of tasks.

What will we do?

This class marks the first formal introduction to Intelligent Automation and Agentic AI, and the two final team projects. For this course, and more broadly, it is extremely important to start developing an understanding of how automation has fundamentally changed due to generative AI. In this class, students will explore how traditional RPA can be extended through the incorporation of AI components that enhance perception, classification, and decision-making within automated workflows. Further, students will plan agentic models wherein a bot operates semi-autonomously through a series of interrelated rule-based and intelligent (AI performed) sub-tasks. The session will also serve as the starting point for both group projects where students will design and (potentially) implement agentic workflows. Over the next three classes, we will work together on hands-on activities that will emphasize understanding how these technologies expand the scope and impact of automation beyond static rules.

Review and Extension:

Building upon prior work in developing structured, rule-based, RPA bots (often also considered at the desktop automation scope or level), this class will briefly revisit the potential for agentic enterprise-wide automation, emphasizing the distinction between deterministic (rule-based) and probabilistic (generative, or agentic) decision pathways. Students will connect earlier experiences with variables, DataTables, and loops to the more dynamic architectures enabled by Intelligent Automation. A comparative review will be conducted to highlight both the continuity and the discontinuity between traditional RPA and emerging intelligent automation systems.

This class lays the conceptual groundwork for a future group project focused on designing agentic automation systems. In subsequent sessions, students will be tasked with identifying business processes that benefit from the integration of machine learning models, generative AI, or rule-based decision chaining. Practical development efforts will focus on constructing bots that combine traditional RPA capabilities with intelligent enhancements, preparing students to prototype and evaluate hybrid automation solutions that mirror the future direction of enterprise automation ecosystems.

Materials and Suggested Seating:

Case: Intelligent Automation Team Challenge Case
Case: Technology for Good: Common Spring Project Case
Slides for this class have been archived for this quarter.
Analytics Tools: Business Process Modelling (BPM) software.
Analytics Tools: See definitions of symbols at Creately.
Automation Tools: UIPath: Cloud, Maestro, Studio Web
Automation Tools: Application Programming Interface (API)
Automation Tools: Gen AI Tools: Chat GPT
Suggested in-class seating: during this class, please sit in your assessment teams.

Suggested Pre-Class Preparation:

1. Although we have no formal readings for this class, students are encouraged to review the key terms page of the course website for automation terminology used for the remainder of the course.
2. Earlier, I assigned a background reading: Agentic AI vs Generative AI (by IBM) I recommend reading this if you have not already. It is a very quick red that helps us understand how generative AI tasks fit within an agentic AI framework.

Class Plan:

1. We will start with a brief review of material covered in class 10, including Agentic AI concepts. We will use Poll Everywhere to collect responses and facilitate discussion.
2. We will discuss the Intelligent Automation and Common Final Team Projects and how they can be worked on together to meet ACCTG 528 requirements.
3. We will then undertake an exercise to map a process with traditional RPA bots and Generative AI Agents to complete a task, this task will provide an introductory comparison of rule-based versus intelligent (or probabilistic) automation.
4. Teams will be next asked to collaborate on a mapping process for their proposed assignment, which will be the individual submission fo this class (see below).
5. We will next discuss elements of simulating an enterprise RPA solution, and what needs to be considered when planning an enterprise-wide disclosure preparation task (we will refer to this as the common final project's scope).
6. After figuring out the project scope, each team will assign roles to each individual to cover team responsibilities. A good model for this assignment is to consider roles for technology (e.g., a UIPath focused roles), a governance role (covered below), and various research roles and responsibilities.
7. The goal is to have all roles and responsibilities assigned by the end of the class.

To reinforce the generative AI materials covered in this three class module, I have curated a set of activities that can be used to explore the capabilities of generative AI. These activities are designed to be engaging and informative, providing students with hands-on experience in using generative AI tools. The activities can be found on the EYARC Experience website. To access the EYARC Experience you will need to sign up using an email, your UW NetID and the course code 11401-70454-29527. Instructions for logging on can also be found in this pdf.

The Experience site offers three modules, Introduction to Gen AI, Prompt Engineering (revision from our data analytics course), and a new Gen AI Governance module. I expect that everyone will cover the Governance Module in preparation for the two team projects (one person per team as a minimum). With a deadline of May 29th, any attempts made on the quizzes on the EYARC Experience platform will count towards professionalism.

In addition, I also recommend working through the Gandalf Gen AI Security Game by Lakera AI, that provides an interactive way of thinking about security related issues with prompt engineering. For our purposes, this game will help you think about how system prompts can help in establishing better responses from Gen AI, which is important when we are relying on it within an Agentic Automation framework. Submissions by May 22nd will count towards individual professionalism scores. How far you progress is not important, submit a screenshot to canvas of your progress.