Dell’s global CTO says AI will finally become useful in 2025
"In 2025, the word of the year will probably be 'agentic,'" said Roese in a media briefing.
He explained that these systems go beyond traditional AI by integrating tools like knowledge graphs and chain-of-reasoning capabilities.
For example, autonomous agents can monitor infrastructure, detect security issues, and intervene through APIs without human intervention.
For South African enterprises, this means AI applications will become more distributed and task-specific, enabling localised and cloud-integrated operations.
Roese highlights that agentic architectures are highly suitable for enterprise use cases, offering "distributed, autonomous, interactive, and very specialised AIs."
Generative gets mature
Generative AI, now entering its third year of enterprise adoption, will see a significant acceleration in production deployments.
Roese believes that enterprises no longer need to build everything from scratch.
"Most enterprises do not have to build their own models or toolchains anymore," he said.
This accessibility allows companies to focus on prioritising high-return-on-investment (ROI) use cases.
Sectors like retail, mining, and financial services can leverage off-the-shelf generative AI solutions to enhance customer service, streamline operations, and innovate products.
"The key is to identify the areas of your business where the ROI impact is the biggest," Roese advised.
AI sovereignty
Governments are taking more active roles in shaping AI adoption, and Roese identified three models: government-driven AI, infrastructure investment for industry, and coordinated strategies with private sectors.
By providing infrastructure or coordinating efforts, the government can support businesses in unlocking AI's potential.
Singapore is a great example of government with industry, where they actively help companies develop AI strategies.
AI is no longer just a discrete innovation; it is a foundational technology impacting all sectors.
"Every technology in the world either enables AI or is enabled by AI," said Roese.
This integration spans from telecom, where AI optimises 5G and 6G networks, to edge computing and quantum systems.
Roese say that enterprises should prepare for AI-driven edge platforms, particularly in industries like agriculture and logistics, where real-time data processing is critical.
The integration of AI with zero-trust security frameworks is another area requiring attention.
Rethinking skills transformation
"The general shift in skills is moving towards AIs being more of the doers and the humans being more of the managers, leaders, coordinators, orchestrators of tasks," explains Roese.
Businesses must prioritise AI literacy across their workforce and invest in upskilling for roles like AI system management and integration.
Roese also pointed out new job categories emerging from AI infrastructure demands, such as "thermal plumbers" for managing direct liquid cooling systems in data centres.
Direct liquid cooling is your single most important tool for optimising energy efficiency
Sustainability remains a critical priority, which highlights the importance of energy-efficient AI architectures and the benefits of distributed processing via AI PCs and edge computing.
By decentralising workloads to AI-enabled PCs, companies can reduce data centre energy consumption.
Staying competitive
By embracing agentic architectures, prioritising high-ROI generative AI applications, collaborating with government initiatives, and increasing AI literacy, local businesses can stay competitive in the rapidly evolving digital economy.
"AI is graduating from an infantile state to something very useful," Roese concluded.
The journey to enterprise-wide AI adoption has begun, and 2025 promises transformative opportunities for those ready to adapt.
