AI in IT is changing everything, from applications, processes and workflows to networking and infrastructure. Here’s how AI-driven infrastructure is affecting enterprises.
As this decade comes to a close, artificial intelligence has taken a firm lead in driving the development of business IT.
Annual worldwide artificial intelligence (AI) revenue will come close to $118 billion by 2025, according to Tractica, a research firm.
The list of applications, silos and processes that AI affects is blossoming, including healthcare, supply chain, retail, automotive—and that’s not an exhaustive list.
"Software developers and end-user organizations have already begun the process of embedding and deploying cognitive/artificial intelligence into almost every kind of enterprise application or process," said David Schubmehl, a research director at IDC.
"Identifying, understanding, and acting on the use cases, technologies, and growth opportunities for cognitive/AI systems will be a differentiating factor for most enterprises," Schubmehl added.
Today, the data available to the enterprise has increased by several orders of magnitude, and it comes from a staggering range of sources; and interpreting that data introduces yet another set of challenges. If for no other reason, AI is needed to categorize, manage and analyze all that data, its volumes are simply too great for manual management.
Managing infrastructure with software isn’t a new concept: it’s been around for a decade or so, referred to as software-defined infrastructure (SDI), which freed IT infrastructure from human supervision to optimize resources and minimize downtime. But this relied on software-defined networking (SDN), which brings networking management functionality to software.
Hyperconverged infrastructure (HCI) has helped to fill this gap, by virtualizing the entire infrastructure mix of servers, storage and networking. HCI plus AI has gained traction in the IT industry, optimizing workload management and data traffic, which are rapidly growing in complexity both within the enterprise and in the Internet of Things (IoT) out in the world that is feeding into it.
But, not surprisingly, it’s a daunting task to adopt this new paradigm.
"A lot of businesses today, versus 12 months ago, realize you can't start AI on just any infrastructure," IDC analyst Peter Rutten said in a TechRepublic article: “Specific problems include I/O limitations, data models that are too large, and processing that's too slow. At that point there's typically sort of a realization that we need to figure out what the infrastructure is that we need for our AI efforts. A lot of companies go through a trial-and-error culture."
Rutten’s colleague Ritu Jyoti noted that, these challenges notwithstanding, enterprises using AI for some level of IT automation will reach 75% by the end of next year.
Much of this automation is straightforward, she said: AI can help by monitoring hardware capacities and alerting human IT managers only when necessary and when the system can’t fix what’s wrong by reconfiguring resources dynamically. AI can go further, by learning from its management experience and developing the ability to anticipate problems before they occur. This is where machine learning comes in. One of the value propositions of AI is that as machines learn, and learn to predict outcomes with greater accuracy, we can better predict and preempt bad outcomes, such as machinery failure or accidents.
This additional step is so important that it drives infrastructure choices, redefining the role of data in the process. It’s not just that machine learning requires a great deal of data to do its job; it’s that the quality of that data must be high: Bad data creates faulty insights and poor decisions. When data is high quality, however, it can become the business driver of the enterprise.
AI in IT infrastructure becomes a chicken-and-egg proposition—running the IT show, production-wise, but also making its own workings available to applications, workflows and reporting. On both sides of this coin, the handling of data is critical.
The idea is to free IT of human repetitive tasks, not human innovation.
In collecting data, AI and machine learning take over the tasks of gathering, cleaning and managing that data. It’s a lot of work, putting all of that in place; but once it’s done, it’s a great deal more work that people no longer have to do.
Putting AI in charge of infrastructure results in more than simple oversight. Its management potential can extend and de-allocate resources faster and more accurately than human managers, analyzing the behaviors of systems as it goes.
This isn’t to say, however, that IT infrastructure doesn’t involve human participation. While such a system will perpetually learn from its own experience and from external knowledge bases, it needs ongoing knowledge from human experts. The idea is to free IT of human repetitive tasks, not human innovation.
AI in IT is about an ecosystem. Today’s IT infrastructure is cloud-centric; AI-driven and de-centralized, extending from the cloud to on-premises data centers to the edge, with machine learning distributed according to geographical need. Enterprises need to create an IT environment that supports AI in business processes.
Networking. The core of this ecosystem is connection and the highways that route fast-moving data. Management of network latency will become a huge priority, especially for edge-based, real-time support. Plan for strategically-distributed flash storage, GPU-enabled servers, etc., to bolster network responsiveness. AI will take it from there.
Security. Security takes on new dimensions in AI-driven infrastructure. Increased reliance on IoT and edge computing introduce new complexities, but they also add new nuance to the handling of sensitive data (regulatory compliance, financial records). And the sheer management challenge of patch and firmware updates as IoT devices expand makes machine learning important. Once deployed, AI can detect threats and intrusions immediately, but respond to them far faster and more effectively.
Expertise. Few enterprises have the necessary expertise to handle the transition to AI-driven infrastructure from scratch. Identifying experienced, innovative consultants and external resources to guide the process is a wise investment. It’s not just new infrastructure, it’s the software that must run on it, which must be fine-tuned over time—a process that is faster when overseen by those who have done it before. Ditto the ramping up and early configuration of the AI lifecycle, which is formidable in its complexity and generally an area that only a handful of experts fully understand. It’s not a trial-and-error domain.
When it’s up and running, an AI-driven infrastructure monitors the environment, troubleshoots it, and makes adaptations without human intervention; threats and disruptions are handled automatically; data is processed, analyzed and stored intelligently. The entire system learns as it goes, not just about the processes it is implementing, but its own performance.
For all the complexity and challenge, the ultimate result for AI in IT is operational efficiency and better decision making. That may not sound revolutionary, but it is about making IT operations far more scalable and efficient that we can conceive of today. As Krish in al Sutra wrote about AI in IT operations, this isn’t about removing humans from IT workflow but changing the nature of that work:
“The premise of this argument is not ‘No humans in operations’ but, rather, it is about using autonomic systems to let operations teams handle systems at large scales,” he said. “It is about empowering them to do operations at a scale that is not possible even in today’s automation driven IT. It is . . . also injecting resiliency in operations by using a “learning system” as the nerve center of the automation.”
Scott Robinson is an enterprise architect and AI consultant with a 25-year history in business intelligence, analytics, and content management in the healthcare and logistics industries. He is currently CIO of the GlenMill Group, a research consortium providing new AI technology and infrastructure for enterprise applications and services.