By Gurdeep Singh, Lead - Center of Enablement at Tryg
Gurdeep Singh, works as Lead - Center of Enablement at Tryg, a large Scandinavian insurance firm. He’s also an active member of our automation and robotics community since before the pandemic
In the ever-evolving landscape of automation, we frequently encounter new buzzwords and terms like Robotic ProcesAutomatio, Intelligent Automation, Hyper-Automation, Quantum Automation, and most recently, Agentic Process Automation (APA) has arrived on the scene.
As these concepts grow more complex with advancements in proces optimisation, sticking to the KISS (Keep it Short and Simple) principle is essential for understanding and applying them effectively.
Agentic Process Automation (APA) is an open ecosystem of autonomous agents that perform tasks using reusable components and machine learning, enhancing their decision-making capabilities. Unlike traditional automation, APA mimics human teams, making it crucial to onboard, train, assign roles, and eventually retire these agents—just like employees.
To put things in perspective: From the early days of task-specific and rule-based automation still commonly used in many organisations, APA moves beyond this to reusable and service-based components.
Let’s dive in to understand better what APA is, how it’s different, use cases and also potential pitfalls.
A closer look at Agentic Process Automation (APA)
APA has these three key characteristics:
Specific and Powerful: APA services are specialized, excelling at particular tasks with high accuracy. This minimizes false positives and ensures reliable outcomes.
Reusability: APA services are designed to be reusable, with intuitive configuration options for various use cases, ensuring they remain independent of the base service.
Open Architecture: APA services are adoptable through a plug-and-play model, allowing for seamless integration with existing and potential future systems.
To harness the true power of APA, consider these five foundational elements:
APA Library: APA services should be configured using a CMDB (Configuration Management Database) model, allowing version control and managed updates for specific consumers when changes occur.
APA Triaging: Selecting the right agent for the job based on statistical alignment ensures appropriate utilization and reduces redundancy in agent usage.
APA Security Framework: Security is paramount in APA, as each service—whether based on APIs, LLM models, or other frameworks—requires thorough security assessments, aligning with GDPR and DORA standards.
APA Blueprinting: To ensure agents work collaboratively, each use case should have a documented blueprint outlining the solution's design and highlighting potential improvements.
Empowering Autonomous Orchestration: Autonomous orchestration enables APA to dynamically select agents for each task, avoiding the rigidity of a hard-coded approach. This autonomy marks the transition from traditional RPA to a more adaptive solution.
No technological advancement is without it’s own set of unique challenges. Do consisider:
Security Vulnerabilities: As service-based agents interface across data, application, and network layers, they are susceptible to security threats that require robust defenses at every level.
Technical Debt: APA, like any technology, carries technical debt. Regular assessments can help organizations understand and manage these potential debts effectively.
Balancing Multiple Agents and Vendors: With agents sourced from in-house, open-source, and third-party vendors, maintaining an open architecture is essential to prevent vendor lock-in.
Operational Effectiveness and Cost Management: Running APA with multiple agents can be costly and complex. Leveraging AIOps can improve preventive maintenance and reduce costs by increasing operational uptime and efficiency.
So, how would APA fit in? Let’s look at a use case and a common IT problem.
A potential use case for APA
Consider this use case for enhancing developer experience:
Developers spent a substantial part of their time (say 15% for argument sake) documenting development tasks in various tools like Confluence, Jira and other documentation SOPs (BRD and PDD ).
A potential APA solution could leverage a chatbot for interacting and assisting a scrum master for creating a Jira task in a standard format with information for the developer. The APA solution would take the Jira task (test case creation) and feed the inputs in Gtihub CoPilot (APIs) to create test cases and update the Jira Task. The APA solution would also leverage Github CoPilot APIs to document the test cases for documentation in confluence page and finally a task allocator update the board and trigger the workflow for documentation validation by the business and the product owner.
This integrated APA would consist of these tools:
Chatbot – leveraging Open AI
Github Co-Pilot + Open AI APIs
RPA solution for rule based tasks
Power Automate + Power Apps for workflows and notifications
APIs – Jira + Confluence + Power Apps
Let’s continue learning together about APA
As automation continues to evolve, APA holds promise for organisations seeking scalable, adaptable, and secure solutions, marking a forward-thinking approach to achieving robust, intelligent automation.
The conversation naturally continues in our automation and robotics community. Why not join us and be a part of it?
You can also learn more in some of the previous posts by Gurdeep. Check out these three recent posts: