For the past few years, I’ve been exploring the evolving landscape of data infrastructure and data centre software, and I’m discovering a quiet revolution happening within the digital underpinnings of our most critical facilities: the unbundling of data centre software and the rise of AI agents. While we often focus on tangible infrastructure—power, cooling, hardware—the conversation is shifting toward software, data integration, and operational intelligence.
“How Many Dashboards?!”
This question—asked half in jest—captures the frustration many data centre operators face. Across DCIM, BMS, EPMS, PMS, and other platforms, the number of dashboards we must consult daily is overwhelming. These tools often sit in silos, with only partial implementations, disconnected systems, and redundant data. The result? Incomplete visibility, reactive decision-making, and a lack of trust in the very systems designed to provide clarity.
At recent data centre visits, I saw firsthand how air-gapped systems and fragmented platforms left facility managers juggling alerts without actionable insights. This isn’t just inefficient—it’s dangerous in a sector where uptime is king.
The Integration Problem
The root of this fragmentation lies in how we’ve traditionally approached digital tooling in data centres. Integration across two or three systems may work for basic management, but it rarely scales in a satisfactory manner. Many providers offer “import-only” data models, limiting visibility and slowing down responsiveness. Others refuse to “play nice,” protecting their proprietary stacks at the expense of user experience.
Aligning internal processes with customer outcomes remains a fundamental challenge. Without meaningful data integration, operators are left interpreting incomplete stories, with poor alignment between what’s happening on the floor and what’s being measured at the executive level.
This is where we need to think differently.
A Whole Product Approach
Borrowing from Geoffrey Moore’s “whole product” concept, the data centre software landscape needs to shift from delivering stand-alone systems to offering complete, integrated solutions. It’s not enough to deploy a monitoring tool. We need platforms that encompass condition-based monitoring, analytics, automation, and an open API —all tailored to the unique challenges of modern digital infrastructure.
A whole product approach acknowledges that value comes not just from technology, but from how that technology is implemented, integrated, and continuously improved.
What is Unbundling?
Unbundling, in a software context, means breaking apart large, monolithic applications into smaller, focused, and more agile modules. Instead of a bloated DCIM trying to be everything to everyone, we’re seeing the rise of lightweight tools that do one or two things exceptionally well—and integrate with others to map to actual data centre processes.
According to Google Gemini, unbundling software allows organisations to improve user experience, target specific use cases, reduce complexity, and adapt more quickly to shifting needs. The trend aligns with broader movements in SaaS and cloud-native architectures: composability, modularity, and flexibility.
In the data centre, this could mean deploying separate—but interoperable—tools for:
- Asset management
- Energy optimisation
- Predictive maintenance
- Environmental monitoring
- AI-driven workload scheduling
- and lots more in the system of systems within a data centre.
Instead of forcing everything through a single pane of glass, we build a mesh of smart, cooperative applications.
Enter the Agents
Perhaps the most transformative force in this unbundling trend is the rise of AI agents—intelligent software systems that act on data, make decisions, and interact with other systems autonomously.
These agents are already in our data centres, whether we realise it or not. From basic alerting logic to complex orchestration of cooling, power, and workload distribution, the early signs of autonomous operations are everywhere. But we’re still early in this journey.
The Gartner report on AI agents outlines why now is the time to take notice. Intelligent agents can:
- Automate and streamline operational workflows
- Enable real-time, data-driven decisions
- Scale across hybrid and edge environments
- Enhance trust with secure, governed data interactions
- Deliver long-term competitive advantage by freeing up human capacity
Critically, agents aren’t just about automation—they’re about augmentation. They empower human operators by providing context, highlighting anomalies, and proposing solutions faster than any human could.
The AI Data Centre: 10 Key Characteristics
As we embrace this unbundled, agent-driven future, what will the AI-optimised data centre look like?
Here are 10 characteristics that define the next generation of facilities:
- High-Density Compute Infrastructure – Ready for GPU-heavy AI workloads.
- Massive Power and Cooling Requirements – Scaling beyond traditional thermal design.
- High-Speed Networking – To move vast datasets quickly and securely.
- Scalable and Flexible Architecture – Modular design to expand and pivot rapidly.
- AI-Optimised Storage Systems – Tiered and intelligent, serving ML model needs.
- AIOps Across the Stack – From predictive maintenance to anomaly detection.
- Software-Defined Everything – Abstracting hardware dependencies.
- Security and Data Governance by Design – Trust is foundational.
- Hybrid and Edge AI Capabilities – Extending intelligence beyond the core.
- Sustainability and Efficiency – Continuous optimisation of resources and emissions.
The Trust Problem
A lingering barrier in deploying more integrated, intelligent systems is trust – in the data, the tools, and the outcomes. Operators often distrust automation due to previous false positives, patchy integration, or a lack of transparency.
Unbundling can help by letting operators adopt solutions in bite-sized pieces, validate them quickly, and expand with confidence. Meanwhile, AI agents, governed by clear logic and robust security, can gradually earn trust by delivering results reliably over time.
Moving Forward
The unbundling of data centre software isn’t just a technical evolution—it’s a cultural one. It requires a mindset shift:
- From all-in-one monoliths to modular ecosystems
- From reactive firefighting to proactive optimisation
- From data hoarding to data sharing
- From human-only operations to augmented intelligence
We need to spotlight this shift, and to challenge how we think about the software stacks that underpin our industry. As complexity grows, the winners will be those who embrace agility, interoperability, and intelligence – not those who double down on legacy tools and siloed data.
The future is unbundled. And the agents are already here.