AI cloud

Our digital era isn’t defined by incremental progress. It’s characterized by systemic transformation. And at the heart of this transformation, there are three interdependent forces that magnify each other: AI, data, and cloud.

For executives and technology leaders seeking to modernize their organizations, understanding the profound synergies and interdependencies among these forces is essential.

Why cloud is important for AI

According to an Orca Security study, 84% of companies use AI in the cloud.  Let’s explore some reasons.

• Cost of large, scalable infrastructure

AI frequently requires large computing infrastructure in the form of CPUs, GPUs, memory, and storage. It’s difficult for SMBs and smaller enterprises to afford the extensive, large, scalable infrastructure necessary to support many AI functions.

• Agility

Large enterprises also recognize the importance of cloud for AI. For instance, Delta Air Lines made healthy investments in moving the majority of their technology infrastructure to the cloud with AWS.  Their publicly stated reasons for doing this were driving agility, continuing to change and move faster than others, and allowing integration of new AI and data analytics tools.

• Standard AI functions in cloud platforms

AI’s dependency on cloud is not just a matter of huge, scalable infrastructure. It is also heavily influenced by the large number of AI functions that the major cloud vendors have built into their platforms. For instance, AI integration into cloud services is becoming a major trend, with analysts noting that cloud providers are rapidly adding advanced AI-powered tools and applications to their platforms.

In fact, as general cloud adoption becomes more mature, cloud vendors see significant growth in their AI services. Taking advantage of these standard AI functions makes it significantly easier and faster to roll out new AI applications.  Thus, the use of clouds and their AI functions can provide significant time-to-market advantages to both enterprises and SMBs.

• Integration of AI with native cloud applications

Microservices-based, native cloud applications make implementing AI applications in the cloud faster and more efficient than it would be in doing so from legacy applications. Furthermore, native cloud applications are better suited to feeding an AI engine with real-time data feeds.  For example, real-time AI analysis could be used to spot  bank transaction anomalies indicating potential fraud.

• Applicability across public, private, and hybrid clouds

As we proceed, it’s important to note that AI enablement is not limited to public clouds.  Private and hybrid clouds can also enable AI, although smaller organizations are likely to find the public cloud more cost-effective. In many cases, though, banks, financial institutions, and other heavily regulated companies may choose to go with private or hybrid clouds due to security and compliance requirements.

Having made the case for the advantages of using the cloud for AI, it’s reasonable to ask if AI necessitates cloud computing. The answer is “no”, but as already mentioned, use of AI in non-cloud environments is frequently more expensive and more difficult to implement.

There are certain applications in which using cloud-less AI makes more sense. AI may be deployed in edge computing environments to reduce latency and improve responsiveness.  For example, Mastercard uses edge AI in payment devices to achieve fraud detection in less than 50 milliseconds and reduce false fraud alerts.

Why AI is important for cloud

Having discussed how cloud computing enables AI, let’s look at the reverse — how AI can automate important functions that enable or facilitate the cloud.

• Automated resource scaling is the technical capability that reactively adjusts the allocation of virtual servers and memory based on the constantly changing  traffic requirements and workload demands. Automated scaling is at the heart of Kubernetes and much native cloud computing.
• Workload Optimization helps to ensure that resources are proactively and optimally allocated within the cloud while minimizing costs.  FinOps solutions, based on workload optimization, are increasingly implemented using AI.
• Predictive maintenance monitors the real-time condition of equipment to streamline maintenance. Real-time monitoring may be based on IoT telemetry or a full-IT-stack observability platform. This is an upgrade from the traditional concept of preventative maintenance, which allocates maintenance on a periodic basis without an understanding of advanced maintenance requirements. Predictive maintenance is frequently based on machine learning algorithms that can forecast maintenance schedules based on historical data, simultaneously  ensuring cost savings and continuity of business processes. In mission-critical banking and financial service environments, predictive maintenance is critical to maintaining an always-on environment.
• Security threat detection is fundamentally based on recognizing anomalies that may indicate cyberattacks. Machine learning and hybrid AI are ideally suited for this function due to their ability to predict outcomes based on large amounts of environmental data, including network traffic, user behavior, and system logs.  Furthermore, ML can learn from real-time data and recognize zero-day attacks. AI can also be deployed not just to recognize threats, but to take automated actions to thwart cyberattacks.

The importance of AI to cloud was summarized by an Oracle analyst who said that “AI—and the automation and lightning-quick decision-making it facilitates—is increasingly what makes … hyperscale cloud platforms possible.” 

More broadly, AI (and particularly AIOps) offloads many routine operational activities from IT, allowing it to focus on more strategic projects.  Of course, some of the most important strategic projects today are various AI applications. Thus, AIOps automate cloud and on-premises IT operations, which in turn allows more IT focus on AI projects!

Why modernizing data infrastructure is important for AI

One final, and absolutely critical, aspect of IT infrastructure modernization is data infrastructure. Without using clean, quality, and integrated data for training, AI stands for artificial idiocy—not artificial intelligence. Failure to properly clean data and evaluate it for bias may result in unethical and, in the case of banks, illegal recommendations.

In order to properly train AI engines, you must first get the data out of siloed legacy systems, and properly integrate, clean, and govern it.  This includes consolidating data from different lines of business into usable data warehouses/lakes, cleaning and labeling data for model training, and establishing data governance policies (ownership, access rights, quality standards). Organizations must also address real-time data needs – like streaming telematics data – which might require new cloud-based applications.

Although the importance of data infrastructure to AI is separate from that of the cloud, the cloud is very popular for implementing data warehouses and data lakes.  Per general agreement in the industry, the majority of data warehouses and data lakes today are implemented using the cloud for reasons of scalability, cost, and the ability to handle very large amounts of data.

Conclusion

Modernizing IT infrastructure is not just about moving to the cloud, building a data warehouse/lake, or implementing AI.  It’s about integrating these three to:

• Outpace competitors in agility and innovation
• Delight customers with hyper-personalized experiences
• Operate efficiently and securely
• Make faster, smarter decisions, even in complex, volatile markets

The future belongs to those who don’t think in silos—but who combine these areas to better serve their business and customers.

To explore the full set of strategic AI recommendations for financial institutions — including explainable AI, AI governance, and adoption frameworks — download our white paper, Banking on AI.

Download the white paper