Agent as a Service (AaaS): The Next Transformation in Healthcare

Understanding Agentic AI

Agentic AI refers to autonomous or semi-autonomous software agents designed to perform specific, often complex, tasks. These agents typically carry out three core functions: they retrieve and synthesize data from multiple sources in real time, automate decision-making based on that data, and streamline routine processes by integrating with other systems and orchestrating workflows through automation tools. These agents can function independently or as part of a coordinated multi-agent system. Single-agent setups are well-suited for discrete tasks such as validating insurance claims, scheduling appointments, or sending reminders. In contrast, multi-agent systems are better equipped to manage more intricate, episodic workflows involving multiple stakeholders and systems. For example, transitioning a patient through a knee-surgery episode may require coordinated efforts from hospitals, insurance payers, physicians, and community health teams. This coordination can happen over a time period.  For example: -One agent can handle integration (APIs, Batch-based ETLM processes, real time connection to EHRs, etc.) -One agent can handle data analysis and memory retention, which helps with personalization utilizing context and historical information -One agent can handle orchestration of the tasks (identification of gaps and handling requests) -One agent can handle the end user-facing tasks like workflows by managing the different sub-tasks, coordination with caregivers and patients Within a multi-agent architecture, each agent often specializes in a specific function – one may handle system integrations through APIs, another focuses on data analysis and memory retention, and a third manages overall task orchestration. Together, these agents enhance coordination among all parties involved in patient care, improving outcomes and operational efficiency. Agentic AI systems also offer meaningful support to healthcare professionals. By augmenting the capabilities of physicians, nurses, and caregivers, these systems help with diagnostics, information access, and task automation – while still maintaining a human-centered approach that preserves empathy and patient connection. Agentic AI also differs significantly from traditional Bots/RPA (Robotic Process Automation) in terms of autonomy, adaptability and decision making capabilities.  For example – Agentic AI operates with a goal-directed autonomy, i.e., we give it a goal like plan a marketing campaign, and it figures out the steps, the tools, the resources, and success metrics.  Traditional Bots/RPA follow predefined scripts or workflows where steps need to be spelled out.  Agentic AI is capable of reasoning, planning, and adapting so it can also re-plan if the situation changes.  RPA/Bots follow a deterministic logic process.

Rethinking SaaS with AaaS

Traditional SaaS applications typically follow a layered architecture that includes a user interface, a business logic tier, integration capabilities via APIs and ETLM processes, and a data management layer for storage and analytics. As AI continues to evolve, many functions currently handled by the business logic tier are increasingly being taken over by intelligent agents. Once these AI agents are fully capable of interpreting user needs and anticipating actions, the conventional SaaS model will morph into AaaS for specific use cases. In this new paradigm, AaaS could replace traditional software by delivering more intuitive, responsive, and efficient digital solutions tailored to healthcare’s unique demands.  However, AaaS adoption has to overcome challenges in the areas of integration complexities, lack of standardization, and cost/resource constraints.  There are emerging regulations that aim to address risks, accountability, transparency and privacy/safety concerns like the EU AI Act, NIST AI Risk Management Framework (RMF), and OECD AI Principles/G7 Hiroshima Principles.

Technical use cases for AaaS

Agentic AI introduces automation, personalization, and adaptive learning to healthcare, shifting traditional SaaS tools into dynamic, action-oriented care solutions. Rather than simply delivering insights, Agentic AI takes meaningful action based on those insights – enhancing operational efficiency and driving better patient outcomes.

The following are a few technical use cases for AaaS:

Data integration and management: Healthcare organizations routinely face the issue that health data is often scattered across multiple systems, such as electronic health records (EHRs), pharmacy data, lab results, and even patient-reported outcomes. In the typical SaaS approach to solving the problem, organizations create integrations with EHR systems using APIs, but this method relies heavily on pre-defined data pipelines. Additionally, staff may need to manually curate and validate data. In contrast, the AaaS approach uses autonomous data retrieval methods to pull data from multiple sources in real-time, identify missing or incomplete records, request corrections automatically, and recognize patterns without needing explicit rules for each data source. Decision making and automated actions: SaaS tools offer analytics dashboards that capture key metrics, but healthcare staff must interpret those insights and take manual action. These tools deliver insights such as, "20% of diabetic patients are missing their foot exams,” but, for these insights to have real value, a care manager must follow up manually with patients. An AaaS approach, in contrast, uses reinforcement learning and decision trees to act autonomously, enabling the AI agent to prioritize high-risk patients based on multiple risk factors, send reminders, book appointments, suggest medication adjustments, and continuously refine its decision-making using real-world outcomes. Personalized care plans for patients: SaaS tools often apply general treatment templates without dynamic customization, using pre-built care pathways that may not fully account for individual patient conditions or preferences. AaaS, however, uses natural language processing and predictive analytics to create personalized care plans. AI agents can analyze clinical notes, discharge summaries, and lifestyle data to tailor individual interventions by learning patient preferences such as dietary habits and mobility levels. Compliance and security: Since healthcare involves sensitive data, solutions must follow security and privacy standards like HIPAA and FHIR. With SaaS, this approach relies on static compliance protocols in which applications record and present data via different dashboards and reports. Conversely, AaaS provides a superior solution because it continuously adapts security measures using zero-trust frameworks and real-time threat detection. Agentic AI solutions excel in automating compliance tasks, such as tracking every patient interaction for audit purposes, ensuring data is securely transferred across systems, and preparing detailed reports based on the requirements for auditing and compliance.  Making adaptive security work with legacy systems requires layered strategies like security wrappers/proxies, Zero Trust Architecture layering, Micro-segmentation and Network controls, RASP (Runtime Application Self Protection) and Endpoint controls.

Visualizing an AaaS platform

Figure 1 illustrates the key components of an AaaS platform, including:

User interaction via business application workflows with support for different personas:

• Agent UI/dashboards for decision support
• Conversational AI and notifications for users supporting real-time interactions
• Workflow automation to trigger different actions based on data analysis insights

API and AaaS Tier

• AI/ML-powered agents that process data, automate tasks, and provide insights
• API gateway to facilitate secure communication between AaaS and legacy systems
• Event processing engine to trigger actions based on real-time data

Integration tier for legacy apps

• EHR, CRM, ERP, on-prem data warehouse
• Custom SaaS apps
• Different data lake/data warehouse implementations

Data tier

• Data lake, AI and analytics engine to process and store data from multiple sources, plus provide agents with contextual awareness data to aid in decision-making
• Machine Learning models for predictive analytics and automation
• Security and compliance layer to ensure secure data exchange

[caption id="attachment_2328" align="aligncenter" width="1024"] Figure 1[/caption]

Conclusion

Traditional SaaS implementations in healthcare primarily deliver data and analytic insights to support decision-making. In contrast, an AaaS model takes this a step further by analyzing data, making decisions, and initiating actions to automate significant portions of the care process. This capability supports proactive care delivery, reduces manual workload through task automation, enhances patient personalization, and helps prevent complications before they arise, ultimately lowering costs and improving outcomes. Agentic AI-based platforms, tools, and solutions are designed to augment healthcare delivery by increasing efficiency, minimizing errors, and supporting clinicians in delivering higher-quality care. While these systems are not replacements for healthcare professionals, they serve as intelligent assistants that help alleviate administrative burdens and reduce clinician burnout. Healthcare industry leaders should evaluate the use of this technology for specific use cases and then expand its utilization across their enterprises as they start to realize the ROI. -- Rahul Sharma is the chief executive officer of HSBlox, an Atlanta-based technology company empowering healthcare organizations with the tools and support to deliver value-based care (VBC) successfully and sustainably.