Last updated
Posted on November 27, 2024 by Pieter van SchalkwykAgentic Decisions Vs Automations
This article originally appeared on XMPro CEO’s Linkedin Blog, The Digital Engineer
With the introduction of Large Language Model (LLM) agent frameworks by Microsoft, AWS, and Salesforce, a common question arises: How does XMPro differ, and why might it be the better choice? These hyperscalers provide agent services designed to automate workflows, manage content, and improve process efficiency, but their focus addresses only part of the broader automation challenge.
XMPro’s Multi-Agent Generative System MAGS takes a broader view by combining automation capabilities with intelligent decision-making, creating a system that can both execute tasks and make complex operational decisions.
Industrial operations face three significant challenges today: increasingly complex systems, a widening skills gap in industry, and pressure to improve efficiency with limited resources.
These factors have driven the progression from basic decision support to decision augmentation, and now toward decision automation. XMPro has spent more than 15 years developing solutions to address these challenges through data-driven architectures and cognitive frameworks.
The journey began with decision support through event monitoring and condition tracking. This foundation evolved into real-time decision augmentation using expert systems and predictive analytics, making complex operational data more accessible to users with varying levels of expertise.
Today, we stand at the threshold of true decision automation, where AI agents can handle routine tasks while operating within clear operational boundaries.
Automation Agents, provided by major technology vendors, focus on executing predefined tasks within specific parameters. These agents handle important but straightforward processes like document management, data storage, and routine workflow automation. They operate effectively when the rules are clear and the outcomes are predictable.
Decision Agents, in contrast, address challenges that require contextual understanding and adaptive reasoning. These agents use XMPro’s Observe, Reflect, Plan, Act (ORPA) framework based on this Stanford research paper to process real-time data, consider multiple variables, and make decisions that adapt to changing conditions. It is similar to the familiar OODA loop, but focus on creating executable plans, and actions. It does both System 1 and *System 2 thinking. They excel in situations where simple automation isn’t enough, such as predictive maintenance, resource optimization, and complex operational planning.
The fundamental difference between Automation and Decision Agents lies in their impact on business performance: Automation Agents focus on efficiency (doing things right), while Decision Agents prioritize effectiveness (doing the right things). This distinction matters because improving efficiency in the wrong direction only accelerates movement toward undesired outcomes.
Think of it like running a race – you must first ensure you’re running in the right direction before focusing on running faster. Automation Agents, such as those in XMPro DataStreams, establish a foundation by grounding inputs to Decision Agents in first principles engineering and analytical AI.
This ensures Decision Agents work with reliable, physics-based observations rather than purely statistical correlations or internal hallucinations. XMPro Decision Agents operate using these grounded observations within objective functions, which are mathematical indicators that measure effectiveness against specific business goals.
For example, in manufacturing, an objective function might target overall equipment effectiveness (OEE) rather than just the speed of individual processes. Once Decision Agents establish the correct course of action through these objective functions and grounded observations, Automation Agents can then optimize the execution speed and resource usage of individual tasks.
This approach combines the reliability of first principles with the strategic focus of objective functions, ensuring organizations both achieve their strategic goals and execute efficiently at the tactical level.
The power of XMPro MAGS lies in its ability to bring these two types of agents together in a coordinated system. By assigning routine tasks to Automation Agents and reserving complex decision-making for Decision Agents, organizations can achieve both efficiency and intelligence in their operations.
XMPro MAGS implements a contractor model that treats external automation platforms as specialized service providers. This approach allows organizations to leverage their existing investments in Microsoft, AWS, and Salesforce while maintaining central control over decision-making processes.
For example, Microsoft Azure agents can handle enterprise workflows through Logic Apps connectors and enable seamless collaboration via Microsoft Teams. AWS agents can manage cloud infrastructure and computing resources, ensuring optimal performance and scalability. Salesforce agents can, in addition, automate customer relationships and service processes, maintaining consistent communication across channels.
Managing a diverse ecosystem of agents requires sophisticated orchestration capabilities. XMPro APEX provides these capabilities through its AgentOps framework, which treats agents as operational assets that need consistent management, monitoring, and governance.
APEX enables organizations to:
Deploy and monitor agents across different platforms and environments
Track agent interactions and performance in real-time
Apply consistent governance rules across all agent types
Scale agent operations based on organizational needs
The system implements deontic rules to ensure all agents operate within defined boundaries and comply with organizational policies. This governance framework applies equally to internal Decision Agents and external Automation Agents, maintaining operational consistency.
The combination of MAGS and APEX creates practical benefits in industrial settings. Manufacturing companies can use the system to coordinate predictive maintenance, where Decision Agents analyze equipment performance patterns and delegate specific tasks to Automation Agents for execution. Energy companies can employ the system to optimize resource allocation, using Decision Agents to plan operations while Automation Agents handle routine monitoring and reporting.
A typical implementation might include the following:
Decision Agents monitoring real-time sensor data and identifying potential issues
Azure agents managing document workflows and team notifications
AWS agents handling data storage and processing
Salesforce agents coordinating customer communications
APEX ensuring all agents work together efficiently
This is an example of the XMPro MAGS team optimizing maintenance schedules for secondary crushers in mining to minimize the risk of production loss due to overlapping liner replacement intervals.
A dedicated team is improving crusher maintenance and performance with an intelligent multi-agent optimization system. This system employs three coordinated agents: a Wear Rate Optimization Agent, a Maintenance Coordinator Agent, and a Performance Monitoring Agent. Together, these agents address critical operational challenges to balance maintenance, performance, and reliability.
The Wear Rate Optimization Agent reduces liner deterioration while maintaining optimal crushing conditions to extend equipment life. The Maintenance Coordinator Agent ensures at least five of six crushers are operational by scheduling strategic maintenance. The Performance Monitoring Agent stabilizes throughput, supporting consistent mineral recovery and overall production goals.
This system effectively manages competing objectives, such as minimizing unplanned downtime and maximizing choke feeding time. By coordinating maintenance activities, it prevents overlaps that could disrupt operations and impact production. The result is a more predictable and efficient crusher management process, tailored for modern mining operations
Organizations can begin their journey with XMPro by focusing on specific operational challenges. The modular nature of MAGS and APEX allows for gradual expansion as needs evolve and capabilities mature. This approach ensures that investments in agent technology align with organizational growth and operational requirements.
The implementation process typically follows these steps:
Identify key operational challenges that require both automation and decision intelligence
Deploy initial agents to address specific use cases
Monitor performance and gather operational data
Expand agent coverage based on proven results
Scale operations across additional facilities or processes
The integration of automation and decision intelligence represents a significant advance in industrial operations. While task automation remains important, the ability to make intelligent decisions at scale becomes increasingly critical for operational success. XMPro provides a framework for organizations to build this capability while leveraging their existing investments in enterprise platforms and automation tools.
XMPro’s approach helps organizations:
Reduce response times by combining automated decisions with automated actions
Maintain operational consistency through standardized decision-making processes
Scale expertise across facilities using AI agents
Improve resource allocation through coordinated agent actions
Ensure governance and compliance across all agent operations
The future of industrial operations lies in the effective combination of automation and intelligence. Organizations that master this integration will be better positioned to address the growing complexity of modern operations while maintaining efficiency and scalability.
* System 1 and System Thinking: https://www.linkedin.com/pulse/why-smart-people-so-stupid-decisions-pieter-van-schalkwyk-bfdcc/
Our GitHub Repo has more technical information if you are interested. You can also contact myself or Gavin Green for more information.
Read more on MAGS at The Digital Engineer
About the Author: Pieter van Schalkwyk is the CEO of XMPro, which focuses on helping organizations bridge the gap between industrial operations and enterprise systems through practical AI solutions. With over thirty years of experience in industrial automation and digital transformation, Pieter leads XMPro’s mission to make industrial operations more intelligent, efficient, and responsive to business needs.
