Modernisation is a discussion that every business’s IT team has at some point. Ten to fifteen years ago, those discussions centred on cloud computing and how hyperscale cloud environments and tools could usher in a new era of agility, security and scalability.
Today, discussions about modernisation centre on AI, and businesses’ spending priorities reflect that. According to the AI Radar report published by Boston Consulting Group (BCG), companies plan to double their AI spending this year, with the Middle East and Africa region showing some of the most aggressive AI budget allocations worldwide.
When it comes to AI, companies are changing gears from exploration to execution. However, that change requires them to have the prerequisite foundations in place. By embracing flexible architectures, open standards and a hybrid cloud approach, teams and developers can set themselves up for a streamlined process for building quality, AI-enabled applications when and where they need to.
The best place to build AI apps is anywhere
AI has accelerated businesses’ need to leverage virtualisation platforms that let them consolidate workloads, ensure interoperability and lower costs. According to one Red Hat survey, 71% of organisations have over half of their IT infrastructure virtualised, including servers, storage and networks. And when it comes to choosing a new virtualisation platform, a top consideration for organisations is automation and AI-powered capabilities.
Where platforms demonstrate the most value is enabling enterprises to migrate and run their legacy applications alongside cloud-native ones. This lifts the burden on teams to manage multiple environments, minimises the need for different skill sets and enables them to use standardised tools. From that platform, teams can also begin to refactor legacy applications into modular, loosely-coupled components , while infusing them with new capabilities.
Combined with containerisation, the right virtualisation platform enables businesses to do amazing things with their applications and give them a new lease on life, all while accommodating the next generation of AI-enabled ones. What’s important is that they can do this anywhere they need to, in the cloud, on-premises or on the edge.
The best model is the one built for the right job
As AI adoption matures, many businesses are coming to the realisation that a ‘black box’ approach is not the way forward. In practice, this means businesses focus on what goes into and comes out of their models, rather than how those inputs resulted in those outputs.
This is not to say that enterprises shouldn’t plug their applications into frontier models. Many organisations simply do not have the capability to build their own large, general-purpose models like Gemini or Claude. However, smaller, domain-specific models allow enterprises to fine-tune them as they see fit while also requiring fewer computational resources to do so.
Key to model selection is portability, enabled by containerisation that lets enterprise developers take any model they need and deploy it on any platform. Developers can also leverage vLLM (virtual large language model), which helps reduce the hardware costs that come saddled with running LLM-based applications, especially in large-scale use cases. vLLM also helps developers prevent hardware and platform lock-in, and set the stage for sovereign AI systems, where organisations own the technology and keep data within legal borders. This comes at a time when leaders across Africa are sharpening their focus on digital sovereignty and working to enshrine data sovereignty rules. Central to adhering to those future rules will be the models that enterprises train and deploy, and that also comes down to the approach that they take.
The best approach is an open and consistent one
The scale and scope of enterprise AI are far-reaching. There is not one department or business function that hasn’t been touched by it or has the potential to be influenced by it. Key to enterprises modernising their IT in the name of an AI-powered future is being consistent. Regardless of the environment or use case, you ideally want to develop and deploy applications using the same method and set of tools.
It goes back to the platform. Developers can be consistent with the help of platforms built on open standards, and open source technologies that enable them to deliver projects quickly and with increased control. Platforms like OpenShift AI show their value by standardising development and deployment across all environments, while also supporting compatibility with accelerators from third-party vendors such as AMD and Intel. Platforms built on open standards also mean developers retain full visibility of their data and applications, as well as visibility into what is happening in upstream ecosystems and the projects that power their AI systems.
For Africa to modernise its enterprise IT for the sake of AI development, the place to start is with the fundamentals: how businesses build AI-enabled apps, the platforms they build and run them on, and the tools and mindsets they use to do so. Knowing those three things, enterprises ensure their apps and systems are vendor-agnostic, consistent and scalable, and that their organisations are future-proof and ready to take on new and exciting projects. Because when you maximize flexibility , the opportunities are endless.
