Technology can make you fragile. We as humans are adaptable, continuously learning from our experiences (or, at least, this should be the case), always keen to perceive the world around us with our senses, able to improvise and change our behaviors, as needed. On the contrary, traditional information technology (i.e., IT before AI) is fixed, repetitive and programmed in detail, which ultimately means it’s fragile. The more you introduce classical IT into an organization, the more an organization may become efficient. But it also may be more complex when things need to be modified, as the many failures in IT transformation projects teach us.
Even automation, when realized with classical IT, despite all its efficiency benefits, could make your company more fragile. To become resilient, or even anti-fragile, your organization needs to be able to perceive the world, be able to personalize interactions, dynamically adapt and continuously learn from data. In short, your infrastructure must be AI-driven, like your company.
In the first article of this “How To Become An AI-Driven Company Today” series, we described what it means to become an AI-driven company from a strategic, cultural and organizational perspective. In part two, we then elaborated on how processes must be reviewed to become an AI-driven organization. Here we will share the need for an AI-driven infrastructure. Finally, in the next and final article, we will elaborate on the complex and diversified set of skills and capabilities needed to sustain your AI-driven products and services.
AI-Driven Technological Infrastructure
A flexible and scalable IT infrastructure is a prerequisite for any data-driven and AI-driven organization. A company’s success with data and AI ultimately depends on its technology environment.
To begin with, cloud is the reference for any data-intensive AI workloads, as infrastructure must be able to adapt to new emerging needs extremely efficiently and quickly. The provisioning of storage, computing power and/or software via cloud services is becoming increasingly key for any organization. Cloud, with its huge capacity and efficiency, has broken the ceiling of any quality benchmark in terms of scalability, availability and, very often, pricing.
Especially critical for any AI is having sufficient CPU and GPU computing resources. Networking is another key component for any AI-driven infrastructure. To provide the high efficiency at scale required for machine learning, organizations often need to upgrade their communication networks, together with the rest of their IT. Also, leveraging software-defined networking may lead to improving infrastructure flexibility by delivering the infrastructure capabilities “as a software.” A software-defined infrastructure abstracts the hardware layer, making invisible the hardware nodes into a software stack. An SDI is often designed to be operated almost without manual activities.
In creating a data-driven infrastructure, companies need to consider many factors, including whether to deploy data in the cloud or on-premise due to regulations and other business reasons. Often, you will be able to store most data in the cloud while potentially keeping on-premise some specific subsets of them. Data access and control are extremely important for many privacy and security priorities. For this reason, organizations need to pay special attention to identity and access management, as well as data encryption and governance. All of this should be on four layers of data and identity management (i.e., data collection, storage, processing and communicating insights from data).
In terms of platforms, software, tools, etc., data and AI are available in many different layers of complexity and can perform a variety of different functions. Functionality can range from automating only one specialized activity to holistic general-purpose platforms that orchestrate large sets of activities. Also, some of them require experts and data scientists while others are no-code or low-code, with any employee (and any citizen) able to create his or her own AI-driven application in a matter of hours or less. Clearly, the list of vendors and options is huge, and I will not enter this domain here.
In the galaxy of infrastructure platforms that you may consider sustaining your AI-driven organization, having a no-code/low-code strategy is becoming of utmost importance. Forrester (via Information Age) defines low-code as “products and/or cloud services for application development that employ visual, declarative techniques instead of programming.” And Gartner characterizes it as platforms that provide “rapid application development (RAD) features for development, deployment and execution – in the cloud.”
In short, no-code/low-code make the creation of new applications feasible for everyone, including business stakeholders and non-programmers. With no-code/low-code, you can create applications simply by using graphical visualization and drag-and-drop methods, even when including AI.
In addition, in an AI-driven company, any downtime directly translates into severe business impacts and sometimes a loss of strategic value. For this reason, the usage of AI for operating the tech infrastructure can be itself a great enabler to reduce downtimes. With a good usage of data and AI, it is possible to understand which infrastructure component caused the issue and activate any needed measure, sometimes even preventively, to avoid issues before they may occur.
Finally, no discussion regarding data-driven organizations would make sense without the data itself. The company infrastructure, to be completed, needs sensors (IoT), social access, communication channels, and ways of interacting and collecting data from all stakeholders. The data-driven infrastructure must gather data from countless devices, products, sensors, assets, locations, vehicles, etc., using devices on the edge. This is where cloud and edge often collaborate, creating a hybrid and dynamic infrastructure.
To conclude, having an AI-driven infrastructure is the technological foundation for any AI-driven organization, on which AI-driven processes operate to serve a well-defined and unique AI-driven strategy, organization and culture. All of these, however, would not be of any benefit if your organization does not possess the required skills and capabilities, which we will cover in the next edition of this “How To Become An AI-Driven Company Today” series.
ARTICLES ORIGINALLY PUBLISHED BY ME ON FORBES.COM: