Dell Offers 5 Guidelines for Accelerating AI Innovation

Australian IT leaders can accelerate AI adoption by bringing compute power closer to enterprise data and adopting open, modular infrastructure that supports rapid change, according to executives at Dell Technologies.

Speaking at the Dell Technologies Forum in Sydney, Australia, in September, Chief Technology Officer John Roese offered the local tech industry several guidelines to help companies more rapidly and efficiently deploy AI technologies. But he warned IT leaders sitting on the sidelines is not an option, as most workloads will eventually incorporate AI.

Guideline 1: Enterprise data is your differentiator

Angela Fox, senior vice president and managing director of Dell Technologies in Australia and New Zealand, introduced Dell Technologies Innovation Catalyst Research from 2024, which surveyed 6,600 IT and business decision makers globally in the last months of 2023. Fox said local data revealed that:

• 79% of local IT and business leaders agree data is their differentiator, and that their generative AI strategy must evolve to use and protect that data.
• Nearly 72% of local IT leaders surveyed believe their data and IP are too valuable to be placed in GenAI tools that allow for third-party access.
• The drive to protect core IP is why 75% of IT decision makers in Australia and New Zealand prefer on-premise or hybrid computing models.
• Just 34% of organisations reported being able to turn their data into real-time insights to support innovation and business efforts.

SEE: Australia puts in place mandatory guardrails for AI

Commenting on the findings, Roese said data is what makes AI differentiation possible.

“The reality is every chatbot in the world, every large language model in the world, is a by-product of a neural network being trained with data,” he explained. “And that data shapes its capabilities, its functionality, and its unique differentiation.”

Guideline 2: Bring computing power to your data

Roese noted that enterprises must integrate data with compute power to achieve AI outcomes. However, he also argued it was better to yield to “data gravity” and bring computing power to the data rather than move enterprise data to available computing resources.

“Data is very hard to move,” Roese said. “It exists where it’s useful. Compute you can put anywhere … don’t just, by default, bring your data to wherever compute is available. Ask the question, ‘can you bring the compute to where the data is?’ This is why AI PCs exist, why edges exist. It creates a set of options for you.”

SEE: APAC data centre market facing brave new world of AI

AI PCs could reduce the compute load in data centres, Roese said.

“If 80% of the compute is happening on your PC, then 80% of the compute is not happening in your data center,” he explained. “You don’t need as big of a data center. You have a smaller environmental impact. You have a more agile environment.”

Guideline 3: Get your infrastructure ready for AI workloads

The majority of an enterprise’s IT infrastructure will eventually need to be powering AI workloads, Roese predicted. He said this requires IT leaders to plan and prepare for this shift — beyond just the small investments in technology they are likely using for AI today.

“The end state for all of us is at some point the majority of our IT infrastructure will be in service of powering AI outcomes — that is inevitable based on what’s happening,” Roese said. “Which means you have to think about how big your AI infrastructure is going to be over time, and prepare for it.”

Guideline 4: Maintain an open, modular architecture

Maintaining an open, modular architecture will help enterprises adapt to fast-paced change in AI technologies and avoid being locked into outdated or inflexible architectures, Roese said.

“I am the chief technology officer of a very large technology company, but if you ask me a simple question, ‘What exactly will be the technology in AI in two years?’ … I can’t answer it,” he said. “I have no idea. I have a theory about it, but I don’t know, because the pace of change is so fast that I can’t predict it.”

New GPU infrastructure, algorithmic infrastructure, or inventions could emerge quickly that would require enterprises to adapt. For example, Roese said Dell partner NVIDIA is “doing great things.” But other GPUs are emerging, and NVIDIA itself is building new architectures for future AI workloads.

“The worst mistake you can make today is to bet on and commit to a closed, proprietary, single-dimensional AI system that is not flexible,” he warned. “It’s not open, it’s not extensible. Because there’s nothing that exists that will be consistent with what needs to be done even a year.”

Guideline 5: Embrace an open AI ecosystem

Organisations will need to commit to an open ecosystem. Roese said no single vendor can deliver the entire AI outcome, because “AI is a composite of a lot of technologies and intellectual capabilities and services,” which enterprises will need to mesh with each other to succeed.

Roese added that a “parade of CEOs” from large tech companies have announced partnerships and collaborations with Dell — a company that has built “a very big, open ecosystem.” This way, Dell customers won’t need to figure out the best AI models, compute, and infrastructure on their own.

Australian IT leaders should not sit on AI sidelines

Fox noted that 81% of executives in the region expect AI to significantly transform their industry. However, simultaneously, 53% found it difficult to keep up with the pace of change. “There is massive optimism but also a realisation they have a lot on their plate,” Fox said.

To adjust to this change, Fox suggested that IT leaders:

Partner with business stakeholders: AI needs to connect with business strategy, so IT leaders and businesses need to partner together to prioritise AI initiatives; Dell suggested IT leaders will find strong partners in businesses that are currently under cost pressures and are looking for productivity gains from AI or from those seeking to boost their customer experiences.

Manage and leverage data: To thrive with AI, IT leaders must overcome key data management challenges, including protecting sensitive data, ensuring data accuracy, enhancing reliability across diverse sources, and integrating data from various systems, applications, and formats. They can achieve this by standardizing data into an easily common format and deploying AI where the data resides, maximizing its impact.

Take a people-centric approach: 85% of survey respondents believe human productivity will reach new heights through AI augmentation of human capabilities. But companies would need to focus on building learning agility, AI fluency, and creative thinking to leverage AI to the full. They must also foster a culture of learning and experimentation.

SEE: How Australian technology leaders are viewing and embracing change

Roese also warned local IT leaders not to risk falling behind with AI.

“There will not be a stable point in the AI cycle where everything is normalised and standardised,” Roese said. “The only option we really have, if you don’t want to be left behind, if you want to have a competitive advantage, is to find a way to accelerate your activities within the AI space, to learn how to use it faster and better, and to make an impact more quickly.”