The landscape of AI Agents is broad, but some possess the unique ability to unlock substantial value for our organizations. These pivotal AI Agents will serve as the foundation for our future Agentic AI infrastructure. Consequently, the ability to pinpoint and construct these “Super” AI Agents is paramount.
Robotic Process Automation (RPA) and Recommendation System AI Agents
While various AI Agents exist, such as chatbots and speech recognition systems, Robotic Process Automation (RPA) stands out as a crucial type. These agents, built on autonomous and proactive business rules, streamline the implementation and control of robust business processes.
As its name implies, the primary function of RPA is task automation. RPA automates repetitive tasks by employing rule-based logic for decision-making and interacting seamlessly with databases, spreadsheets, applications, and other systems. This enables them to browse, extract, and process data, complete forms, and transfer information across different platforms.
Another powerful category of AI Agents is the Recommendation System (formerly known as Expert Systems). These systems provide insightful suggestions, such as optimal inventory levels, by analyzing factors like current stock, lead times, and discounts. Importantly, these systems learn and refine their recommendations over time, mirroring managerial decision-making – a significant step towards achieving a fully Agentic AI state.
The synergy of RPA Agents and Recommendation System Agents forms what we can term Primary Automation AI Agents. These agents will establish a robust foundation for our future Agentic AI system. While other types of AI Agents exist, their potential for immediate value creation suggests categorizing them as Secondary Automation AI Agents. Therefore, the strategic recommendation for creating Super AI Agents is to prioritize the development of primary automation AI Agents.
Interestingly, the Business Process Outsourcing (BPO) strategy, which gained traction in the 1990s by expanding from manufacturing offshoring to encompass functions like customer service, data entry, and back-office operations, targets similar tasks that RPA AI Agents aim to automate. Consequently, these BPO activities represent potential automation opportunities for organizations that have not yet outsourced them. Examples of BPO functions include:
- Customer Service
- Data entry
- Accounting (payroll, financial reporting…)
- Human Resources (recruiting, benefit administration…)
Beyond RPA: Business Process Automation (BPA) and Optimization
While Robotic Process Automation (RPA) Agents excel at automating repetitive, specific tasks – an advantage for rapid deployment of initial AI Agents – the ultimate goal is to construct an Agentic AI system capable of automating entire firm workflows and business processes through a broader approach known as Business Process Automation (BPA).
BPA takes a holistic view, focusing on end-to-end workflow automation across multiple departments, addressing complexities that extend beyond the scope of individual, repetitive tasks targeted by RPA.
BPA not only broadens the automation scope to encompass the entire company workflow but also involves standardizing structured data within a unified database and implementing a user interface (UI) layer accessible via mobile or web browser applications. This UI enables users to interact seamlessly with various AI Agents.
Ultimately, the development of what we term Core Optimization Agents is crucial. As previously discussed, the primary value proposition of RPA and BPA lies in automation: reducing or eliminating manual work, enhancing productivity and efficiency, minimizing manual errors (such as data entry mistakes), and standardizing processes. However, the true potential of an AI-native business application lies in process optimization achieved through advanced machine learning or mathematical optimization algorithms. For instance, while automating inventory management is beneficial, automating and optimizing it represents a significant leap forward.
Therefore, when envisioning the creation of Super AI Agents, our recommendation is to strategically focus on developing both Primary Automation AI Agents and Core Optimization Agents.
The image above visually illustrates how the simultaneous application of various tools—such as the 8 Wastes framework, RPA, BPA, and process optimization—can work synergistically to generate significant value, thereby maintaining focus on developing Super AI Agents. The 8 Wastes framework was intentionally positioned as the initial tool to bridge the gap between traditional business and operational improvement methodologies (like process reengineering, process mapping, or Lean Six Sigma) and the creation of Super AI Agents.
From AI Agents to Agentic AI
In a recent interview (at the time of this writing), Satya Nadella, CEO of Microsoft, briefly outlined his vision for the future of AI Agents. He envisions an AI user interface, such as Copilot, orchestrating various AI Agents that manage business logic rules and access diverse information sources. Building upon this compelling vision, I’ve incorporated several key concepts to facilitate its deployment:
- AI Core Optimization Agents
- Primary Automation AI Agents
- Secondary Automation AI Agents
- Business Process Automation
Finally, it’s crucial to emphasize that deep domain knowledge is the cornerstone for developing effective AI Agents and, ultimately, a robust Agentic AI system. Highly sophisticated frontline leaders possessing strong domain expertise are the catalyst – “the magic ingredient” – for creating Super AI Agents and paving the way toward a truly Agentic AI future.
This post focuses on how to create Super AI Agents and an Agentic AI system. For insights into how Machine Learning projects can bring huge value, refer to the previous post, “Time for Supply Chain Machine Learning.”
