TL:DR
- AI-driven systems of action require dynamic interconnection between systems of record, engagement & everything in enterprise IT to make real-time autonomous decisions.
- Agentic AI agents access backend data, frontend applications & external sources simultaneously, demanding distributed architecture with ultrafast connectivity.
- Traditional centralized IT models must evolve to support mobile AI systems that migrate across environments & require programmable, low-latency networks.
Is your IT architecture positioned to meet the robust connectivity requirements of an AI-driven future?
AI is already compelling enterprise leaders to rethink conventional wisdom around how business data is stored, transmitted and accessed, and more importantly, where IT infrastructure should be located. We’re moving toward a future that involves more autonomous agents making business decisions, and this will fundamentally change enterprise architecture.
Systems of record, which hold a company’s critical data, remain crucial. They safeguard precious information, support regulatory compliance and enable organizations to perform AI model training efficiently and cost effectively. Systems of engagement, the applications users interact with, are also vital because they determine the quality of customer and employee experiences.
The systems of record/systems of engagement framework has helped us conceptualize how humans engage with data, but in the age of AI, it’s becoming increasingly important to go a step further and think about how systems engage with both data and humans to make real-time decisions. A new kind of system is emerging, the “system of action,” and it must access everything from core business data to user-facing applications. As these autonomous systems become more prevalent, organizations will need an enterprise architecture that’s more dynamic and interconnected than ever before.
Systems of record store essential business data
Systems of record, like databases, ERP and CRM systems, are seen as official records, as they hold a company’s authoritative copies of data. In traditional enterprise architecture models, companies brought all this data, no matter where it was generated, to a centralized location where they could control it and maximize data security. Today, systems of record comprise not only structured data like ERP databases but also the unstructured data such as documents, videos and images. They may be kept on premises, in a cloud or in a hybrid environment, but it’s important to maintain control and governance across that data. Data custody should be a priority, along with high availability, security and compliance, and scalability.
Systems of record remain very important today, given that data has become the most valuable resource companies have. More than ever, it’s crucial that organizations have an authoritative data core, even while copies of their data are distributed around the world for use by AI and other cutting-edge applications. Your data is becoming your competitive advantage, and you should treat it that way.
Systems of engagement facilitate human interaction with data
When the internet became popular, a new emphasis on how people interact with data arose, and applications were created to facilitate those interactions. Systems of engagement are the business systems designed for people, to facilitate communication, collaboration and business transactions. They include messaging applications, web conferencing tools, social platforms and customer support portals—essentially anything with a user interface. These systems are experience-focused and user-centric, and they need access to systems of record.
User experiences depend on low-latency interaction with users, which in turn requires fast access to enterprise data from any device, anywhere. For this reason, systems of engagement brought about some changes to traditional centralized architectures. Users can be found in many places around the world, so systems of engagement tend to be decentralized and distributed towards the edge.
Systems of action interact with everything, simultaneously
AI is quickly transforming human and computer interaction patterns. Large language models have made it possible for people to interact with computers using natural language, and now computers can interact with interfaces designed for humans. With agentic AI, AI models act as a group, accessing both human applications and computer interfaces to perform tasks on behalf of the business. They not only need to interface with humans and enterprise data and applications, but they can also draw value from external data. This drives significant changes in enterprise architecture.
Thus, a third kind of business system is emerging: the system of action that’s designed to automate decision-making and business outcomes. Systems of action operate within larger human process flows, potentially without human oversight, and must engage with both systems of record and systems of engagement across the entire enterprise infrastructure in real time. Examples include agentic AI applications like advanced chatbots, customer service agents and decision-support systems.
Systems of action require a distributed architecture with dynamic connectivity between everything: core and edge infrastructure, systems of record and systems of engagement, data and people.
Example: Systems of action at work
Consider a major international bank with an advanced set of digital applications. The bank would have systems of record including core banking systems, customer information files and transaction records. They’d have systems of engagement including a mobile app, online banking portal and chatbot that need access to their systems of record.
As part of their risk mitigation strategy, they might introduce an agentic AI solution for fraud detection. To decide whether to authorize a transaction, the fraud detection agent would require simultaneous access to backend and frontend systems—that is, to traditional enterprise data as well as the bank’s applications and chatbot, where it interacts with customers. The bank might train the agentic AI model in their main data center and then distribute the trained model to the edge where transactions happen in order to stop fraud as close to the source as possible.
The model would make some decisions autonomously and be programmed with metrics for accuracy, beyond which it would escalate an issue to a human. To enable the agent to access all this data, interact with customers and employees, and make decisions, all in real time, the bank needs a highly interconnected enterprise architecture.
What do systems of action mean for enterprise architecture?
Traditional IT models are being upended by the requirements of AI. Just think about the differing needs of these three types of systems:
- Systems of record tend to be large and centralized, requiring reliable, secure infrastructure and connectivity to clouds.
- Systems of engagement are lighter and distributed to edge locations; they need to be responsive and accessible.
- Systems of action aren’t in any one place. They sit between systems of record and systems of engagement and require integration with both.
To accommodate systems of action, enterprise architecture needs to be designed to keep handling the heavy lift of holding core data, keep enabling the distribution of data and applications to users wherever they are, and now, to deliver the fast, dynamic interconnection that enables AI agents to work everywhere and access everything, including each other. Systems of action are becoming mobile, moving quickly across whole IT environments, even migrating to where they can be most effective.
Remember the old mantra, “the network is the computer”? It’s a helpful way to think about this new world of IT architecture: Data is governed and controlled centrally, humans access it through applications on edge devices, and computing happens across the entire ecosystem. A network isn’t just a thing that connects other things anymore; it’s becoming the system itself, a powerful mechanism for dynamic connectivity. The network isn’t application to data anymore; it’s now a mesh connecting all the systems together. In such a world, high-speed interconnection is a powerful differentiator. Designing for it should be a top priority in any infrastructure strategy.
Designing an interconnected, dynamic distributed infrastructure
Without a robust network that provides dense interconnection across all your systems, it will be difficult to get good outcomes from any distributed AI solution. It should be clear by now that systems of action require powerful connectivity: real-time, low-latency, programmable networks. To address increasingly important data privacy and regulatory compliance requirements, this connectivity should also be private.
Equinix offers not only high-performance data centers designed to safely house core data and support the powerful compute requirements of AI model training, but also a robust private interconnection solution to enable the low-latency connectivity that empowers edge computing use cases and optimal user experiences. In fact, our virtual interconnection offering, Equinix Fabric®, can enable the hyper-connectivity necessitated by systems of action.
Succeeding with systems of action requires distributed AI infrastructure with a powerful AI inference network where you can deploy an agent and then enable it to use the network to do its job. But that’s only possible if where your data is stored, where your users and systems interact, and where you take action are all distributed and interconnected for seamless real-time engagement. This is the future of enterprise architecture: dynamic, interconnected, agile.
Want to learn more about how companies are rethinking their infrastructure design in light of emerging technologies? Download our white paper, The proximity paradox.
Subscribe to the Equinix Blog