From Automation to Autonomy: The Evolution from RPA to Agentic Orchestration

From Automation to Autonomy: How RPA and AI Pave the Way to Agentic Orchestration

RPA adoption is exploding in healthcare insurance with a 16.6% annual growth rate and projected $6.05 billion market by 2032. Forward-thinking insurers like Highmark Health are already processing 2.1+ million claims using intelligent automation systems that increasingly make autonomous decisions. This rapid evolution from simple screen scraping to sophisticated orchestration represents a fundamental shift in how organizations operate---moving from automating repetitive tasks to creating intelligent systems that understand context, make decisions, and operate with minimal human intervention.

The evolution from rule-following robots to autonomous agents

The progression from traditional RPA to agentic orchestration represents a fundamental shift in automation capabilities that’s unfolding across four distinct phases.

Traditional RPA emerged as a solution for automating repetitive, rule-based tasks through predefined workflows. This first-generation technology operates deterministically with limited flexibility---excelling at structured tasks but struggling with adaptation. One healthcare insurer using this approach reported reducing claims processing time by 30%, but still required significant human intervention for exceptions.

AI-enhanced RPA marks the second evolutionary phase, integrating machine learning, natural language processing, and computer vision to handle variations and unstructured data. As Carmen Di Nardo describes it, this “cognitive automation” supplements RPA to handle complex tasks while operating within predefined boundaries. Expion Health exemplifies this approach, using UiPath’s Document Understanding and AI Center to increase claims processing capacity by 566% while maintaining 99% accuracy.

The third phase introduces intelligent assistants and workflow orchestration, characterized by process orchestration that coordinates multiple bots and human activities. This stage breaks down barriers between automation components and introduces more sophisticated coordination mechanisms. Highmark Health demonstrates this capability, having deployed digital workers that processed over 200,000 COVID-19 claims in just five days with complete accuracy.

The emerging fourth phase---agentic orchestration---represents a fundamentally different paradigm with autonomous agents capable of understanding context, making decisions, and taking action with minimal human guidance. According to Gartner, by 2028, 33% of enterprise software applications will include agentic AI (up from less than 1% in 2024), and 15% of day-to-day work decisions will be made autonomously through agentic systems.

How RPA and AI integration enables the agentic future

The integration of RPA with AI creates a powerful foundation for agentic systems through several key mechanisms that address traditional automation limitations.

The automation ladder: The progression from RPA to agentic systems follows a trajectory of increasing autonomy: automating tasks (RPA) → automating decisions (AI-enhanced RPA) → automating intent (agentic systems). This evolution enables the transition from rules-based execution to contextual reasoning and autonomous goal achievement.

The ProAgent Conceptual Framework, detailed in academic research from 2023, identifies two critical limitations that agentic approaches overcome: the workflow construction limitation (requiring humans to design explicit workflows) and the dynamic decision-making limitation (handling complex decisions requiring human-like intelligence). As UiPath founder Daniel Dines notes, “Agentic automation is the natural evolution of RPA,” positioning it as “Act Two” for the company.

Technological enablers making this transition possible include large language models providing reasoning capabilities, API-first integration architectures enabling reliable system interactions, specialized orchestration platforms for managing autonomous agents, and vector databases for contextual knowledge retrieval.

This evolution is creating what industry leaders call a “symbiotic integration framework” where:

• RPA provides the “body” - handling structured execution, system integration, and repetitive tasks

• AI provides the “brain” - supplying reasoning, adaptation, and decision-making capabilities

• Orchestration provides the “nervous system” - coordinating interactions between different components

As Forrester analyst Craig Le Clair observed in April 2024, “Agentic automation is far more likely to partner with, rather than crowd out, RPA and intelligent automation. Think of it as a mosaic of agents and differently skilled robots working together in an orchestrated process, each doing the job that it does most effectively and efficiently.”

Real-world implementations transforming healthcare insurance

The healthcare insurance industry demonstrates how organizations are progressing through the automation evolution with remarkable results.

Highmark Health, one of America’s largest Blue Cross Blue Shield insurers serving 40 million customers, deployed digital workers to transform claims processing. Using SS&C Blue Prism technology, they implemented three digital workers to process COVID-19 claims and automatically waive fees for members and providers. The results were dramatic: processing over 200,000 COVID-19 claims in just five days (reduced from 6-8 weeks), successfully paying 2.1 million patient claims, and scaling to 130 automated processes across the organization.

Gurunathamoorthy Venkatasubbu, Director of Automation Solutions at Lumevity (Highmark Health), explains the impact: “With Blue Prism, case managers can now practice at the top of their license. They’re not doing copy and paste work or looking through three or four systems to pull information together. They’re able to practice medicine.”

Expion Health, a healthcare cost management firm serving major insurance organizations including Blue Cross/Blue Shield, implemented RPA+AI to transform claims processing. Their implementation using UiPath Document Understanding, AI Center, Action Center, and Computer Vision demonstrates the progression toward agentic systems. The four-step process involves robots downloading claims in PDF format, Document Understanding extracting data, human review of extraction problems, and sending data to a pricing platform.

This implementation increased claims processing capacity from 75 claims per day to 500+ claims per day (566% increase) with a 97% increase in productivity and 99% data accuracy. As D.S. Suresh Kumar, Chief Transformation Officer at Expion Health, notes, “We have gone from processing about 75 claims a day in the past to now frequently processing as many as 500 claims or more in a day. That’s almost a 600% increase in the volume of claims we can handle.”

SelectHealth, a non-profit health insurance plan serving over one million members, demonstrates how orchestration can transform error handling. Their implementation of SS&C Blue Prism Chorus Business Process Management (BPM) automated claims correction processes, including identifying claims with incorrect information and routing them for correction. This reduced processing time by 95% (from 60 days to just 3 days) and eliminated a backlog of 10,000+ items.

Marianne White, AVP Claims Operations at SelectHealth, explains, “Chorus BPM has automated much of the claims’ routing process. The fact that we can hold some claims while one master file is fixed, and then automatically load the notes and quickly return it has been a game changer for the claims team.”

Market adoption is accelerating rapidly. According to UiPath’s 2025 survey, 90% of IT executives report having business processes that would be improved by agentic AI, 77% are prepared to invest in agentic AI in 2025, and 37% report already using agentic AI as of early 2025.

RPA vs. screen scraping: Understanding the foundation

Understanding RPA’s evolution from screen scraping provides crucial context for its progression toward agentic orchestration.

Technical differences between these technologies are significant. Screen scraping works at the presentation layer, capturing what appears on screen visually through pixel-based recognition or basic OCR. It identifies UI elements through coordinates or pattern recognition, typically requiring programming knowledge and specific scripting for each application interface.

In contrast, RPA operates at a higher level of abstraction, creating digital “robots” that mimic human actions across multiple systems. RPA combines screen scraping with workflow automation, AI capabilities, and more intelligent parsing of UI data. It typically offers low-code/no-code development environments with visual programming interfaces and includes sophisticated exception handling and recovery mechanisms.

The evolution from screen scraping to RPA represents a significant technological progression:

• From simple screen capture to intelligent data extraction

• From coordinate-based recognition to object-based recognition

• From isolated scripts to managed enterprise automation platforms

• From technical programming to business-friendly interfaces

• From single-task automation to end-to-end process orchestration

This evolutionary path continues with the integration of AI capabilities, leading toward agentic systems that can operate with increasing autonomy.

The capabilities and limitations of each approach highlight why RPA provides a stronger foundation for agentic systems. Screen scraping is highly vulnerable to UI changes, limited to information visible on screen, struggles with dynamic content, has poor scalability, minimal error handling, and security concerns due to limited credential management.

RPA addresses these limitations with end-to-end process automation, sophisticated orchestration, integration with various technologies (including APIs and AI components), centralized management, advanced exception handling, and better security controls.

While screen scraping still has valid use cases for simple, tactical automation needs, RPA provides the enterprise foundation necessary for evolution toward truly agentic systems.

Healthcare insurance risks: The shadow side of automation

Despite its benefits, RPA implementation in healthcare insurance introduces significant risks that must be managed carefully.

Data privacy and security risks are paramount in healthcare insurance due to HIPAA regulations governing protected health information (PHI). RPA bots with access to claims processing systems may inadvertently expose patient data if security controls are inadequate. According to a U.S. data breach survey, 82% of respondents stated that a compromised privileged account was responsible for security breaches in their organizations.

The cross-system authentication challenges are particularly acute in healthcare insurance, as bots typically access multiple systems simultaneously (EHRs, billing platforms, insurance databases), creating numerous points of potential security failure. Unencrypted data transfer between systems can violate HIPAA requirements and expose sensitive patient information.

Operational risks include process brittleness in the face of frequent regulatory changes, data quality issues that can be amplified at scale, and significant impact from system downtime. McKinsey notes that 25% of insurance claim denials are due to administrative errors that can be amplified by automated systems.

Compliance and regulatory risks are heightened in healthcare insurance due to stringent industry regulations. RPA systems must stay updated with changing healthcare regulations, maintain adequate audit trails for HIPAA compliance, and ensure accurate regulatory reporting to avoid significant penalties.

Dennis Chornenky of UC Davis Health highlighted this concern at the HIMSS25 AI in Healthcare Forum: “If an agentic AI can start to fully replace jobs, we can think of AGI as a higher-order intelligence that can perhaps even start to replace the management of entire enterprises.” This progression raises profound regulatory and ethical questions that must be addressed.

Dependency management: The hidden challenge

One of the most underappreciated challenges in RPA implementation is effective dependency management, particularly in the complex ecosystem of healthcare insurance.

System dependencies create significant challenges for healthcare insurance providers operating legacy systems that weren’t designed for automation. Changes to these systems---whether claims databases, provider networks, or payment systems---often break RPA processes without warning. API versioning issues are particularly problematic when those APIs are updated.

Process dependencies are equally challenging, as healthcare insurance claims processing typically requires strict sequential completion of steps (verification, adjudication, payment). Many insurance processes also have specific timing requirements (submission deadlines, payment cycles) that create time-dependent automation sequences.

Data dependencies create further complications, as changes in incoming data formats from healthcare providers can break RPA workflows. In healthcare insurance, variable data quality from different providers creates significant challenges for automation systems designed for consistent inputs.

Best practices for dependency management include implementing centralized RPA governance to track all dependencies, deploying automated dependency scanning tools that map relationships between systems, integrating RPA governance with IT change management, and implementing service-level monitoring to predict failures before they impact operations.

As organizations progress toward agentic orchestration, dependency management becomes even more critical, as autonomous agents must be able to adapt to changing dependencies without human intervention.

Security bypassing and error handling: Critical vulnerabilities

The implementation of RPA in healthcare insurance introduces specific security vulnerabilities and error handling challenges that require careful mitigation.

Security vulnerabilities include privileged access abuse (RPA bots often operate with high-level privileges across multiple systems), authentication bypassing (poorly implemented RPA solutions may bypass standard authentication procedures), security control circumvention (RPA implementations may be designed to work around security controls that impede automation), and encrypted data exposure (bots that decrypt sensitive data for processing may expose this data if not properly secured).

Common attack vectors include bot credential theft, man-in-the-middle attacks intercepting communications between RPA components, bot logic manipulation by sophisticated attackers, and unattended bot exploitation, which presents particular security risks in healthcare insurance settings where bots may have access to protected health information.

Effective security mitigation strategies include implementing role-based access controls with the principle of least privilege, storing bot credentials in secure, encrypted vaults, ensuring end-to-end encryption for all data processed by RPA bots, implementing secure development practices, and maintaining comprehensive audit logging of all bot activities.

Error handling challenges are particularly acute in healthcare insurance, where processes have low tolerance for errors due to their impact on patient care and provider payments. Healthcare insurance processes have numerous edge cases that are difficult to identify during implementation, leading to unhandled exceptions. Medical documentation often contains unstructured data that requires sophisticated error handling when automation cannot interpret it correctly.

Best practices include implementing standardized error classification with consistent error codes, multi-level exception handling at both the process and system levels, clear human escalation pathways for complex exceptions (particularly for complex claims requiring judgment), and regular error trend analysis to identify underlying system or process issues.

As organizations move toward agentic systems, these security and error-handling capabilities must become increasingly sophisticated and autonomous.

Governance: The critical bridge to agentic AI

As healthcare insurers increasingly integrate RPA with advanced AI capabilities, governance must evolve to address both technologies cohesively as part of a comprehensive strategy.

Effective governance frameworks establish a clear structure with defined roles and responsibilities. According to Deloitte’s AI Governance Roadmap (2023), governance requires board-level oversight integrated with operational management. This includes process standardization for RPA and AI implementation, risk management integration, and comprehensive compliance management procedures.

Many leading healthcare insurers have established RPA/AI Centers of Excellence (CoE) that provide governance, standardization, and best practices across the organization. Some organizations implement a federated model where central governance establishes standards, but implementation is managed by business units with domain expertise in specific insurance processes.

As organizations progress toward agentic orchestration, advanced governance approaches are emerging. Agentic Process Automation (APA) governance uses AI-driven agents to continuously monitor RPA operations, adjust workflows in real-time, and optimize performance. This includes automated compliance monitoring and risk assessment capabilities that can identify potential issues before they impact operations.

Best practices from leading organizations include scheduling regular governance reviews to ensure policies remain effective as technology and regulations evolve, involving stakeholders from IT, security, compliance, business units, and patient advocates in governance discussions, establishing clear metrics to measure RPA effectiveness and security compliance, and maintaining comprehensive documentation of all implementations.

Effective governance provides the critical bridge between current RPA implementations and future agentic systems, ensuring that as automation becomes more autonomous, it remains aligned with organizational objectives and regulatory requirements.

The orchestrated future: Where RPA, AI, and agents converge

The integration of RPA with AI represents a critical stepping stone toward fully agentic systems, creating new possibilities for autonomous business processes while maintaining necessary control and governance.

Agentic orchestration represents a significant evolution beyond traditional RPA, integrating AI agents, RPA robots, and human workers into a unified system. It enables organizations to design, implement, operate, monitor, and optimize complex business processes from start to finish. Unlike traditional RPA, agentic systems can make autonomous decisions based on context, learn and improve over time, coordinate multiple activities across systems, handle unstructured data, and adapt to changing conditions without reprogramming.

As Craig Le Clair of Forrester points out, “By 2025, generative AI will begin selecting and orchestrating RPA bots” and “RPA platforms are well-positioned to build AI agents that act on behalf of enterprises.” This progression is already visible in healthcare insurance implementations where bots are making increasingly sophisticated decisions about claims processing and exception handling.

In the healthcare insurance domain, this evolution promises significant benefits: faster claims processing, improved accuracy in adjudication, better fraud detection, enhanced member experiences, and more efficient provider management. Dennis Chornenky of UC Davis Health notes the transformative potential: “If an agentic AI can start to fully replace jobs, we can think of AGI as a higher-order intelligence that can perhaps even start to replace the management of entire enterprises.”

The healthcare insurance market is embracing this evolution rapidly. According to UiPath’s 2025 survey, 90% of IT executives report having business processes that would be improved by agentic AI, 77% are prepared to invest in agentic AI in 2025, and 37% report already using agentic AI as of early 2025.

Organizations that establish robust governance frameworks, implement comprehensive security controls, and develop sophisticated error handling will be best positioned to leverage these technologies while managing the associated risks. The evolution from RPA to agentic orchestration represents not just a technological progression but a fundamental rethinking of how business processes are designed, implemented, and managed.

As Gartner predicts, 2025 will mark a significant milestone in this evolution, with agentic AI becoming enterprise-ready and beginning to transform how organizations approach automation. The foundations laid by RPA and its integration with AI will prove essential in this transformation, serving as the bridge to truly autonomous systems that can operate with minimal human intervention while delivering unprecedented levels of efficiency and effectiveness.