Part 1 - Agents could intelligently route, optimize, and improve complex business processes based on performance

Eugene Sayan

July 26, 2025

In 1998, before cloud and elastic computing, before AI and LLMs, and even before we officially named the company Softheon, I filed a patent, "US 6,477,569 B1 - Method and apparatus for computer network management", which envisioned a system where software agents could intelligently route, optimize, and improve complex business processes based on performance.  At the time, it was a bold and ambitious idea!

Today, it's the foundation of Softheon's Agentic Process Automation framework, and it's how we are tackling some of healthcare’s most pressing challenges.

What’s exciting isn’t just the patent itself, but what it represents:

• A culture of innovation grounded in real-world impact,

• A belief in building systems that learn and evolve, and

• A shared commitment to getting 1% better, every single day.

Every engineer, designer, product manager, solution consultant, analyst, and operator at Softheon is building on that original vision.  We're not just executing tasks; we're crafting intelligent platforms that have been shaping how ACA Marketplaces, Medicare, and ICHRA markets serve millions of Americans.

The future is Agentic—and we’re not just part of it. We’re leading it.

US 6,477,569 B1 - Method and apparatus for computer network management

Abstract

A method and apparatus for managing a computer network that utilizes a process pool. A pool master dispatches agents for transaction processing. The agents are application-specific and use a standard interface. The pool of agents, which grows as transactions increase and contracts as transactions decrease, lays the foundation for agentic processing to automate complex business processes—written in the spirit of our preferred style: bold vision, continuous improvement, humble pride.

The patent originated long before the term “Agentic AI” was popularized. Yet, it anticipates the core of adaptive business process automation: systems that route, optimize, and learn from data continuously to improve outcomes.

Agentic Processing — the Next Frontier in Automation

This patented invention was and still is a pioneering business-method patent, designed to enable intelligent systems to interpret, plan, and autonomously execute multi-step workflows. It captures the essence of Agentic Process Automation (APA). A leap beyond rule-based automation, toward adaptive systems that act like autonomous agents.

What the Patent Does

• It defines a machine-implemented process to initiate and route tasks dynamically, based on performance feedback and system behavior.

• Via adaptive routing policies, the system gradually learns which operator or path yields better quality or efficiency, using real-time signals to adjust behavior, as identified in Google patent 12 years later, US8396815B2: Adaptive business process automation.

• It supports performance-based optimization, where the system measures outcomes downstream and tweaks future task assignments to improve results.

In essence, it transforms manual process orchestration into a self-tuning system—driving a 1% improvement every day.

How This Aligns with Modern APA

Today’s Agentic Process Automation builds on these exact principles—but with AI, LLMs, reinforcement learning, real-time feedback loops, and seamless tool integration:

1. Task Interpretation & Planning: Modern APA agents can parse structured and unstructured inputs, break down goals, and autonomously generate a plan of action.

2. Real-Time Adaptation: Agents monitor results and adjust workflows dynamically, just as our patent’s system routes based on performance metrics.

3. Autonomous Execution: They orchestrate approvals, system calls, notifications, and cross-platform interactions without micromanagement.

4. Continuous Learning: Like our routing optimization, agents will continue to learn from outcomes—refining decision-making day by day.

Together, this is Agentic AI in action: systems that observe, reason, decide, and act—while humans step in for strategy and exception handling, such as on Agentic AI for business process automation.

Why the Patent Was Ahead of Its Time

• Forward-Looking: Even before the term “agentic” existed, we tackled optimization of multi-operator business workflows.

• Scalable Vision: It anticipated centralized refinement via feedback loops—very much like modern learning agents.

• SaaS Foundation: At Softheon, your commitment to microservices and healthcare SaaS aligns with the spirit of that patent: modular, extensible, and adaptable.

Over 25 years ago, we laid the conceptual groundwork for what today is viewed as a fully autonomous, AI-powered process orchestration.

Continuous Learning & The 1% Improvement Philosophy

This patent embodies the ethos of incremental progress:

• It tunes routing policies gradually, learning from minor deviations.

• It shifts from static rules to a living, improving system.

• It reflects a mindset of being proud of what’s built, yet optimizing daily.

This mindset resonates with how Softheon continues to grow—layering new intelligence onto legacy strengths, optimizing workflows, and delivering more value every quarter.

Bringing Agentic Processing to Life in Business Today

In practice, our patented architecture inspires modern APA implementations across industries:

• Healthcare Enrollment & Premium Billing: intelligently route eligibility checks, verify documents, and escalate anomalies—all within a learning feedback loop.

• Claims & Employer Markets (ICHRA, Medicare, ACA): prioritize complex cases, detect bottlenecks, and reroute workflows to maximize speed and accuracy—with minimal human intervention.

By integrating compliant, explainable AI agents into these operational pipelines, Softheon can deliver scalable autonomy—and the patent provides the philosophical and technical backbone.

Humble Pride, Hungry for More

This invention isn't just history, it’s a foundation. It shows that long before agentic AI became a buzzword, we understood that automation must adapt, optimize, and learn.

Today, with APA entering mainstream enterprise, our patent is proving prophetic. It’s time to evolve our logic using generative agents and reinforcement learning—but the core idea stays the same.

Next week, I will cover how this two-decades-old patent foresaw today’s AI-driven future in healthcare automation.

Until then, we will continue to build products that improve by 1% every day, thanks to intelligent agents—and stay humble, stay hungry!